Issue link: https://iconnect007.uberflip.com/i/1440051
JANUARY 2022 I SMT007 MAGAZINE 95 ality of electronic components trade. For ex- ample, contemporary counterfeit components can be mixed with authentic ones in variety packs and avoid detection by sampling screen- ing techniques. Another example is that com- ponents in the inner part of a reel package have different humidity exposure than the reel's out- er part 13-16 . Solderability Assessment Method We classify the solderability of the leads by learning how the solder-lead of a component reflects the light vs. how a lead with poor-sol- derability component does. e light reflec- tion can be performed by analyzing the com- ponent image and/or analyzing its reflectance spectra in the UV/Vis band. In addition, the component's apparent age, can be estimated based on the gradual degradation of the lead's reflectance. is information is available by examining the surface of the soldering leads. We have found that visual analysis of the component image using artificial intelligence methods can detect the degradation of the soldering leads that leads to poor solderability. In order to train a model that correlates the image of the leads and its solderability, we need to design a neural network based on multiple images of leads. Solderability Estimation Algorithm e solderability is estimated by deep learn- ing methods performed on the images of the soldering leads of the components. First, we use the manufacturer year code to calculate the manufactured age of each item. We use this data to train a model so it can predict the items' age with the least prediction error. It is important to note that the manufactured year does not directly reflect the real quality of the lead. is is due to differences in stor- age and handling conditions that may cause accelerated aging to the soldering leads, thus, degrading the "age" of the leads. In addition, there is a distribution to the conditions of the leads in both leads with good and poor solder- ability. e network is designed to fit a linear re- gression model so that the distance from the input dataset to the line is minimized. e next step is to use the regression model to predict the age of components from the validation set. e components are sorted by the estimated or predicted age and then we choose a split point (age value) to split it into two groups. e splitting point is at least three standard deviations between the two groups. e solderability is evaluated by using a tier algorithm. e first tier evaluates the bulk uni- formity of multiple components from the same transport package (reel or tray) using an unsu- pervised approach. e second tier classifies each lead to a group of manufacturing years using a classification algorithm. e third tier classifies each lead to two extreme classes of solderability state: good or poor solderability using a classification model. Solderability Analysis A non-destructive mass volume method for assessing the solderability of electronic leads based on deep visual inspection (DVI) is pre- sented. e method allows real-time assess- ing of all assembled components. It may be de- ployed during the SMT mounting process and by a reel-to-reel incoming inspection. e solderability is correlated to surface re- flectance and degradation in solderability caused by corrosion and intermetallic reac- tions in the surface of the leads. is may be The network is designed to fit a linear regression model so that the distance from the input dataset to the line is minimized.