Research On Chemical Information Related Problems Based On Soft Sensor

With the continuous development of computer science and chemical industry,chemical informatics arises at the historic moment. In the field of Chemical informatics research, chemical industry has always been one of the most important part. In the actual industrial process, the process of data is extremely complicated,there are many unstable factors such as multivariable coupling, parameter timevarying, strong nonlinearity, large lag, etc. These unstable factors make the traditional sensor cannot get effective application in practical application, not be able to obtain rapid accurate measurement of important variables, the production process do not get effective optimization and diagnosis. Soft measurement technology is a new kind of intelligent detection technology, which is widely used in industrial process control system and has a broad application prospect. This paper studied the problem of soft measurement modeling method under the background of smelting gas sulphuric acid concentration in the dry absorption section of test process. Aiming at the existence of the traditional soft measurement modeling method in the small scope, such problems as low accuracy, robustness, in order to solve these problems,this paper do the research about the soft measurement modeling method based on relevance vector machine, and successfully establish the predict sulfuric acid concentration of soft measurement model, the simulation results demonstrate the effectiveness of the proposed method.This paper introduces the basic principle of soft measurement technology,analyzes the purpose and significance of research of soft measurement technology,summarized the research results and development of experts and scholars in China and abroad. The paper also analyzes the application machine learning algorithms. As the dimension of the data is high, this paper adopted a method of attribute reduction to simplify the quantity of the data. Based on this, puts forward a kind of soft measurement model of optimal relevance vector machine. The simulation resultshows that the soft measurement model established by using this algorithm is convenient and effective, accurate prediction precision. In order to improve the accuracy of soft measurement model in sulphuric acid concentration and efficiency,to use "vector" more effectively in relevance vector machine, the characteristics of the low proportion of proposed a rapid relevance vector machine soft measurement modeling method. On the basis of the combination of the the threshold coefficient and maximum limit, reducing the time of the model significantly and put out a new method of measuring the on the basis of combining using iterative estimation of participation of training data to estimate quickly, and reduce the large number of unrelated vector in the training data, to narrow the scope of the training sample,greatly shortened the time of model training, and establish the soft measurement model for the prediction of relevance vector machine. The optimized algorithm is demonstrated by the simulation results in prediction accuracy and time efficiency of sulphuric acid concentration has the satisfactory results.