Research On TFT-LCD Quality Prediction Method Based On Mearching Learching

With the continuous development of artificial intelligence technology and the improvement of industrial requirements,various machine learning and deep learning technologies are gradually emerging in the industrial field."German industry 4.0","made in China 2025" and other slogans have also once again accelerated the development of industrial technology.In tft-lcd(thin film transistor liquid crystal display)quality detection process,the original use of sampling detection method to estimate the quality of the entire sample,but this method is not comprehensive and does not have timeliness.With the rapid development of industrial big data and the continuous improvement of industrial product quality requirements,this simple quality detection method has been unable to meet the increasingly stringent industrial quality requirements,and better industrial intelligent prediction technology urgently needs to be invented.In view of the above problems,this paper analyzes the tft-lcd quality prediction data.The range of labels and the characteristics of data sets are analyzed.The method of data preprocessing and the way of selecting effective data are studied,and encoder coding conversion is carried out for different types of data.In view of the characteristics of small sample size and high latitude of the data set,the traditional dimensionality reduction method was studied,and PCA method was used for feature extraction of the data.In the quality prediction stage,the classification of traditional machine learning algorithms such as supervised and unsupervised learning methods are studied.By combining literature search and experimental comparison,the prediction algorithms with better performance,such as random forest and SVM algorithm,are studied in depth.On this basis,a new regression prediction method "RAN-SVM" algorithm was designed by applying the idea of random forest in decision tree to SVM algorithm,which reduced the sensitivity of single SVM algorithm to data.Finally,the RANSVM algorithm was combined with the traditional random forest algorithm,and the two models were fused using the idea of residual fitting,and compared with the decision tree,KNN,linear regression and other algorithms in the traditional algorithm.The results after the fusion of RAN-SVM and random forest effectively improved the accuracy and robustness of the whole sample,and the results after the prediction using this model were significantly lower than the mean square error of other models.The good performance of the new research algorithm in experimental data also provides a new thinking direction for the development of machine learning algorithm and expands the space for the application of random forest idea in traditional algorithm.Applying this algorithm to the prediction of tft-lcd,the final quality result of the product can be obtained more quickly and accurately,rather than the final inspection of the product through the traditional sampling survey.The use of new prediction technology can be based on pre-known results to make corresponding decisions and strain,responsible for customers,more sensitive to manufacturing production,improve efficiency,improve product quality,improve customer satisfaction.