| With the increasing demand of consumers for electronic products,electronic products are developing towards the direction of miniaturization and refinement.Surface Mount Technology(SMT),one of the core technologies in the electronics manufacturing industry,is facing great challenges.Statistics show that 60%~70% of surface mount quality problems are caused by solder paste printing process,so it is especially important to optimize the solder paste process parameters.At present,the optimization of solder paste printing process parameters mainly has the following problems: it is necessary to obtain regular data through experiments;it is not possible to dynamically recommend printing process parameters in real time online;the data is not fully utilized,and expert experience is not integrated into the analysis.In view of the above problems,this paper proposes a method for optimizing the process parameters of solder paste printing based on data mining.The main research contents are as follows:(1)The overall research framework for optimization of solder paste process parameters is constructed.The SMT process flow and solder paste printing process flow are combed,and the solder paste printing mechanism is analyzed in detail.Based on this,a process parameter optimization method based on data mining technology is proposed.Firstly,data preprocessing and feature engineering processing are performed,and then the prediction model of solder paste quality is built and optimized.Finally,the objective function of solder paste printing process parameters is built and optimized.(2)Analyzing and mining of key influencing factors of solder paste printing quality.In view of the characteristics of large volume and high dimension of solder paste printing data,feature engineering is used to extract key information that has great influence on printing quality.On the one hand,the selection model of solder paste printing characteristics is established to screen the important characteristics that have great influence on solder paste quality.On the other hand,combined with the characteristics of printing data and the experience of process experts,two new characteristics of the scraper pressure per unit speed and the category of PCB lengthd are reconstructed,which provide support for the accurate prediction of solder paste quality.(3)Constructing a solder paste quality prediction model based on particle swarm optimization to optimize deep neural network.Aiming at the complex and changeable solder paste printing process and the correlation of printing parameters,a prediction model of solder paste quality based on deep neural network is proposed,and the standard particle swarm optimization algorithm is improved.The improved particle swarm optimization algorithm is used to optimize the initial weights of the deep neural network model,which improves the convergence speed of the model and improves the prediction accuracy of the model,and provides support for the subsequent optimization of solder paste printing process parameters.(4)Multi-objective optimization of printing process parameters.Aiming at the different contribution weights of solder paste volume mean and volume standard deviation to the printing quality of solder paste,a process parameters optimization objective function with weight coefficient is proposed,using the "data driven" technology to determine the weight coefficient of the objective function,using the improved particle swarm algorithm to optimize the optimal process parameters.This method makes full use of the characteristics of the data itself to calculate the weight coefficient of the objective function,avoiding the blindness of manually setting the weight of the objective function.Based on the above research,the optimization analysis of solder paste printing process parameters is completed,the optimal printing process parameters are solved,and the defect alarm rate is reduced by multi-batch production,which verified the effectiveness of the proposed method. |
PDF Full Text Request