A Method For Predicting The Roughness Of PCB Hole Wall Considering The Characteristics Of Micro-drilling Wear

In the process of printed circuit boards'(PCB)mechanical drilling,the abrasion of micro-drilling will aggravate deterioration of hole wall roughness,which will affect the circuit performance of PCB.In order to control the roughness of hole wall within the standard range in time,it is necessary to evaluate the roughness of hole wall according to the abrasion characteristics.However,in the actual production,instead of strict and accurate quantitative forecasting method,the micro-drilling abrasion and hole wall roughness are only evaluated by the empirical value of drilling times.Therefore,in order to control PCB's hole wall roughness,the key is how to extract the value of micro-drilling abrasion,and then,establish an effective evaluation method of hole roughness according to this characteristic value.As a result,the research purpose of this paper is to extract the number of abrasion of PCB micro-drill,by introducing a series of machine vision algorithm.GBDT algorithm will also be used to evaluate PCB hole wall roughness,which is based on the abrasion amount characteristics,and establish an accurate and effective evaluation method of hole wall roughness.The main contributions of this paper are as follows:(1)Exploring the extraction of abrasion characteristics,in this paper,the actual diameter,the abrasion area of the edge surface and the notch depth of the edge surface were determined,which can reflect the characteristics of the micro-drill abrasion.In order to extract the actual diameter,a machine vision detection method is proposed,it is based on contour superposition algorithm and key point selection algorithm.This method can effectively avoid the influence of cutter grooves,and hence improve the extraction accuracy of the actual diameter features of micro-drill.Aiming at the extraction of PCB micro-drill edge surface abrasion area and edge surface notch depth,an adaptive micro-drill edge surface recognition algorithm was applied,which is based on region growing algorithm.By comparing with OTSU algorithm,this method significantly improved the accuracy of extraction of edge surface notch depth and abrasion area feature.(2)Selecting eigenvalues for PCB hole wall quality assessment.This paper firstly used the survey method,which can be used to obtain all factors that will affect the hole wall quality under actual conditions.The analysis method is to select the eigenvalues,which could be used for quantitative analysis and modeling,and then eliminated the secondary factors for hole wall quality assessment modeling through the grey correlation analysis.In addition,by comparing the evaluation results of the new feature based on abrasion quantity and the traditional feature based on drilling number,and these two kinds of data would be applied to a variety of machine learning algorithms.The parameter mainly based on micro-drilling abrasion quantity was selected as the characteristic value of the hole roughness evaluation.(3)Raising an evaluation model of PCB hole wall roughness considering new characteristic parameters and its system implementation.In this paper,basing on the new characteristic quantity,an evaluation method of PCB hole roughness is proposed.On the model establishment,Bayesian algorithm was applied to optimize the combination of super parameters in the GBDT network.By comparing with the evaluation results of the traditional network,it was found that the prediction accuracy of the optimized GBDT network was higher,which could meet the requirements of the actual production application.In addition,by building a simple graphical interactive interface,a PCB hole wall roughness simulation system based on micro-drilling abrasion detection is completed.