Research On Quality Analysis Model Of Infrared Thermal Imager Based On Data Mining Algorithm

Product quality has become the fundamental guarantee for manufacturing enterprises to continue to maintain competitiveness in the increasingly competitive global market,the government clearly put forward the "strengthen quality brand building" as a strategic task for the transformation and development of manufacturing enterprises,for infrared thermal imager manufacturers such as precision equipment manufacturing enterprises,the importance of product quality is self-evident.With the establishment of enterprise production information system,the data related to product quality has increased sharply,and how to find valuable information from complex data to help enterprises improve product quality and enhance core competitiveness has become an urgent problem to be solved and has great research value.Aiming at the quality data analysis needs of KI Company,an infrared thermal imager manufacturer,Thesis takes the quality inspection data of infrared thermal imager as the research object,and uses data mining technology to construct the quality classification prediction model and association rule analysis model of infrared thermal imager products,classify the quality level of products and mine the potential knowledge related to product quality in the quality inspection data,so as to provide a basis for KI company’s product quality improvement and production process optimization.The main work of Thesis is as follows:(1)In order to solve the problem that the original quality inspection data is disorganized,the format is not standardized,and the data mining modeling cannot be directly carried out,Thesis establishes the quality inspection database to store data,uses interpolation algorithm,data balancing algorithm,feature selection algorithm,etc.to process the quality inspection data of Infra CAM-A thermal imager,and proposes a general process and method for quality inspection data processing of infrared thermal imager,which can process the quality inspection data of different models of products.(2)In Thesis,several mainstream machine learning algorithms and convolutional neural network algorithms are selected to establish quality classification models,and genetic algorithms are used to optimize the hyperparameters of the machine learning models,but the prediction performance of each model is not good.Therefore,based on the principle of model fusion,Thesis proposes a Stacking-LR＿SVM＿CNN-LR(S-LSCL)two-layer fusion model based on Stacking algorithm,which gives full play to the advantages of each algorithm,breaks through the prediction performance limit of single model,and evaluates the model from the aspects of overall prediction performance and minority prediction performance,Experiments show that the S-LSC-L model has better classification performance than other single-model and fusion models.(3)This Thesis adopts the improved Apriori algorithm to establish an infrared thermal imager association analysis model.In response to the shortcomings of the Apriori algorithm in repeatedly scanning the database and generating too many frequent items,combined with the characteristics of low data volume and complex data in the infrared thermal imager quality inspection dataset,an improved Apriori algorithm based on hash algorithm,transaction compression,and dataset compression is proposed.The accuracy and excellent performance of the algorithm are verified through experiments;Using the improved Apriori algorithm to establish a quality inspection data association analysis model,mining strong association rules between various detection items of infrared thermal imagers and product quality,interpreting the strong association rules,converting them into meaningful knowledge,and providing support for enterprise decision-making.