Research On E-jet Printing Accuracy Prediction Method Based On Machine Learning

With the advantages of simple structure,high printing resolution and a wide range of printing materials,electrofluidic jet printing is a new type of micro and nano printing technology that has been developing rapidly in recent years.However,the development of electrofluidic printing has been hindered by the many factors affecting the accuracy of electrofluidic jet printing,the complexity of the electrofluidic forming mechanism,and the low printing accuracy.Therefore,research into the accuracy of electrofluidic printing has been carried out to address the problem of low accuracy of electrofluidic printing.Firstly,the connection between electrode voltage,solution flow rate,nozzle structure and collection plate height and the stable formation of the jet was discussed by means of finite element numerical analysis,and the jet diameter when the droplet reaches the collection plate was used as the accuracy evaluation index.Then,the jet diameter data obtained from the finite element analysis is used as the sample set;finally,a machine learning algorithm is used to build a prediction model and the influence factors are optimally validated by weighted linear fusion of the machine learning algorithm with the NSGA-II optimisation algorithm.The main findings and conclusions of this thesis are as follows.:(1)Explore the law between the main parameters such as electrode voltage on the stable forming of the jet.First of all,the basic principles of electrofluidic printing are discussed,and the fluid level changes and jet patterns of the electrofluidic injection process are summarised.The flow field,electric field control equations Maxwell stress and interface,tracking equations are discussed to provide the basis for subsequent numerical finite element analysis calculations.Then,a finite element model of the E-jet printing is built and the correctness of the finite element analysis is verified in terms of the charge distribution and the shape of the jet,The laws between electrode voltage,solution flow rate,nozzle structure and collection plate height on the stable formation of the jet were investigated.(2)Established an electrofluidic jet printing accuracy prediction model and analyzed and compared related research methods.Experimental data from the finite element analysis of cone jet printing was used as the dataset to build a prediction model for the accuracy of electrofluidic jet printing using linear regression,support vector machine,random forest and neural network machine learning algorithms.Using the decision score as the performance evaluation metric,the support vector regression and neural network algorithms were able to achieve above 0.9 on the test set.However,the linear regression algorithm scored only around 0.7,indicating a low model confidence level.Therefore,non-linear regression methods are superior to lr regression methods.(3)A method for the predictive optimisation of electrofluidic jet printing accuracy is investigated.Firstly,a finite element analysis model for electrofluidic jet printing is developed;on this basis,different machine learning algorithms are used to develop prediction models for each input and output value,and a weighted linear fusion method based on a genetic algorithm is proposed;finally,a method based on a combination of machine learning and NSGA-II algorithms is proposed and used to optimise the electrofluidic jet printing model.This paper investigates the problem of predicting the accuracy of electrofluidic jet printing based on machine learning,which improves the prediction efficiency by more than ten times and saves the time of accuracy prediction.The conclusions of the study are of reference value and practical significance for the selection of electrofluidic jet printing process parameters.