Research And Implementation Of Comprehensive Evaluation System For Tobacco Leaf Quality

Based on the requirements of tobacco company's comprehensive evaluation of tobacco raw materials.After studying various evaluation business needs in the field of tobacco leaf quality evaluation and principles and characteristics of a lot of multi-index comprehensive evaluation methods and models,the software system for tobacco leaf quality evaluation business is designed and implemented.The system consist a Three-Layers software framework which is consist of User Interface-Layer,Business Logic Layer and Data Access Layer based on the.NET environment,which is work together with the Python scientific computing language module to complete the function of the comprehensive evaluation of tobacco materials.The main work completed is as follows:(1)Principal Component Analysis method is adopted,based on the variance within the data set,for the characteristic that the contribution of tobacco sensory quality indicators to the comprehensive score is difficult to determine.The sensory score is evaluated,and sorted them in the sample.(2)The Fuzzy comprehensive evaluation method is used to statistic and analysis the chemical composition indexes of tobacco leaves,the evaluation results can describe the “absolute position” of the quality of tobacco leaf samples in the tobacco quality system.(3)Because of the Gray system comprehensive evaluation model has the characteristics of “small sample modeling”,it is used to establish the model of some chemical component indicators that the value range have not been proved while the character has been ensured,and get the comprehensive score by calculating the “Gray Relative Value” between the sample value and the best value of the set,which is maximize the use of the information already available at this stage.(4)Using BP-Artificial Neural Network to train the training set composed of chemical composition indicators and sensory scores and using the network model which is trained to predict the sensory score,which is,base on the existed data and experiences,release the problem that to obtain network parameters establishing a sensory index system that requires a lot of human resources and time cost through manual evaluation and statistics.(5)Drawing scatter figure to interview the result of some evaluation method to make the result more intuitive..The system can support the tobacco leaf quality index set preserved by varioustobacco companies in the current file format such as Database and Excel which is provided by tobacco company.By testing 492 sensory sample sets of 2010-2013 which sampled form Yunan,Hunan,GuiZhou province,and 92 chemical composition sets of 2015 consists of 111 components,the evaluation result is approved by the tobacco professions.The system selects appropriate evaluation models based on different evaluation business characteristics that is deployed for the tobacco company to deal with the evaluation work whose evaluation results have certain reference and contrast effects for subsequent manual evaluation.