Research On Estimation Approaches For Radiated Emissions From PCBs With The Unsupervised Learning

With the development of Machine Learning(ML),more and more fields can be combined with ML.The unsupervised learning algorithm,which is a part of ML,is good at dealing with massive data.It has an important role in processing unlabeled data.In the field of electronic engineering,PCBs are used as the basis for layout design with the increasing demand for the high-level electronic equipment.The reason of PCBs being the important part of electronic engineering field is that PCBs can greatly reduce the cost of equipment and assembling.Radiation in PCBs could affect the normal operation of subsystems or other systems.Detecting the distribution of radiation sources in the PCB could effectively help the PCB solve the radiation problem.Combining unsupervised learning algorithms with radiation problems in the PCB could optimize the method of detecting the distribution of radiation sources in the PCBs.The main work of this thesis is as follow:Firstly,the technical advantages of detecting outliers by Gaussian distribution algorithm,Self-Organizing Map(SOM)Neural Network and K-means clustering algorithm(K-means)are analyzed.A method for detecting the distribution of radiation sources in the PCB based on SOM and K-means algorithms is proposed.In the beginning,Gaussian distribution is introduced to extract outliers,and then SOM is used to do the first clustering process.After that,K-means is used for doing the second clustering process.Finally,the number of radiation resources in PCBs is determined by the elbow method.The simulation result shows that the proposed method could effectively detect the distribution of radiation resources without detecting radiation distribution by experience.Secondly,researching a way to detect radiation sources of PCBs radiation sources distribution when the data dimension is too high.A method for detecting the distribution of radiation sources in the PCB based on SOM and GMM is proposed.In the beginning,the isolated forest algorithm is introduced to extract radiated data in the high-dimensional data,and then a model composed of SOM and GMM is used for clustering process.Finally,the number and distribution of radiation sources in the PCB is determined by silhouette coefficient method.The simulation result shows the proposed method could effectively achieve detecting the distribution of radiation sources in PCBs.This method could also solve the problem that sources could not be manually judged withhigh-dimensional data.