Development Of Soft Measurement Integrated System Based On Data Driven Technique

With the development of science and technology, soft measurement in industrial soft sensor is involved in the form of various engineering application. Usually soft sensor is mainly applied to deal with the specific problem, determine the soft sensor solution, make the model more targeted, supply more single species model. At the same time, by providing a priori condition guarantee, this kind of soft sensor usually have determined auxiliary variables and data preprocessing method. In practical application, the fixed soft sensor form is hard to adapt to environment flexibly and cannot transplant to deal with such other similar issues. Soft sensor development costs much, the development cycle is long, has low rates of resource utilization and the operator is hard to manipulate.Soft sensor system is integrating soft demanding of engineering sensor system, designing versatility of soft sensor integrated-system software in the case of variable auxiliary variables. This article collate based on the extensive analysis of soft sensor data preprocessing, auxiliary variable selection, research and application of modeling, model calibration and developed integrated system maintainability of software sensor with the higher universal and real-time based on Visual C++6.0. The main work includes:1)Needs analysis of soft sensor integrated-system, functional designAccoding to the principle of soft sensor system, common steps, timeliness and versatility, the dissertaion makes a need analysis for soft sensor integrated-system and sets up the main function of the system in detail.2) Development and realization of soft sensor integrated-systemAccoding to the design basis of soft sensor, build the system framework; Taking into account the online soft sensor’s easy operation and system safety requirements, the system is divided into the soft sensor platform and online soft sensor platform, and realize the development of general soft sensor integrated system based on Visual C++6.0development platform. According to the system’s effectiveness and data security requirements, using the OPC client obtain the real-time data, using the singleton pattern to do the data operation; to achieve the purpose of general soft measurement, using the principal component analysis method as the free choice of secondary variables selection guidance, design the module of data preprocessing schemes, model are respectively provided commonly used methods of soft measurement modeling and multi model fusion, user can construct the model according to the requirements, in order to achieve universal purpose, and under the Visual C++6.0environment call OpenCV library function realize the matrix operation to reduce the complexity of the model algorithm; to ensure high accuracy prediction output model, provide PD and adaptive cross validation algorithm model to correct; Finally, package the involved algorithm, and develop the related modules of the online soft sensor system.3) Test analysis of soft sensor integrated-systemThis dissertation has a software test analysis for soft sensor integrated-system. Accroding to the testing procedure of commen software which are unit tsets, integration test, confirmation test, system test,this dissertation has a comprehensive test for the each module of the system and a priori guarantee for its application.