Research On Analysis Method Of Influencing Factors On The Quality Of SMT Products Based On Big Data

As the core process of electronic assembly,Surface mount technology(SMT)directly affects the quality of SMT products.With that development of economy,the market demand for electronic product is increasing.The layout of PCB board is more complicate,which makes solder paste printing more difficult.The relationship between printing process parameters is complicate,the printing performance is difficult to control,and the defect rate is not easy to reduce.The existing analysis methods of influencing factors on the quality of solder paste printing depend too much on the previous experience,and the analysis of various parameters involved in the production process is insufficient.Moreover,defect rate and performance indicators reflect printing quality together.The existing analysis methods fail to study them at the same time,and the analysis is not comprehensive enough.With the automation and intellectualization of production equipment,the influence factors on printing quality are concealed and more difficult to look for.To solve the above problems,this paper establishes the analysis model of influencing factors on SMT printing quality according to the characteristics of big SMT data and the logical relationship between parameters,and proposes analysis methods of influencing factors on the quality of solder paste printing for different targets.The main work is as follows:(1)This paper introduces the SMT process and big data characteristics briefly,and analyze the mechanism of solder paste printing.On this basis,this paper proposes the analysis model of influencing factors of SMT printing quality,including big data processing,influencing factor similarity measurement,data packet construction,analysis of influencing factors and result evaluation.(2)An analysis method of influencing factors on SMT solder paste printing performance based on random forest feature selection is proposed.The key factors affecting the printing performance got through comprehensive analysis of five printing performance indicators.(3)The solder paste printing defect dataset is high dimensional and less classified information.This paper proposes an improved multi-layer feature extraction and classification method to tackle the problem.The MDS is able to reduce the dimension initially and to keep the sample spacing unchanged as much as possible.Then,using the entropy method to select the main features.The optimal classification surface got by LDA helps to complete the second dimension reduction and classification.(4)An analysis method of influencing factors on solder paste printing defects based on SVMRFE is proposed.Combining with mutual information theory and cross-validation to obtain the optimal number of features and the most relevant factors on printing defects.Finally,the quality improvement suggestions are given.The experimental results show that the above analysis methods of influencing factors can effectively obtain the key influencing factors of solder paste printing performance and defects,and the multi-layer feature extraction and classification method can effectively achieve dimensionality reduction and defect classification.