Study Of Thin-film Evaporator Process Parameter Based On Rough Set Theory

The thin-film evaporator is a new type of highly efficient evaporator. It is used and popularized widely in industry such as chemical, light industry, pharmacy, environment protection and food industry because of its highly heat-transfer coefficient (large evaporation intensity), good effect of low-temperature evaporation, short residence time and wide viscosity applicability. In the study of data collection and inspect system of thin-film evaporator, large amount of process data such as temperature, flux and pressure are produced and accumulated. How to utilize this information effectively and enhance production efficiency through this information has become the problem cried for being solved. Through the data mining of these process data, knowledge of parameter optimization and forecast method of output and evaporation intensity were found. Therefore, the engineering operation can be directed. The process of thin-film evaporation is highly complex. So its information structure is highly non-linear and strongly correlative. These bring much difficulty to the forecasting of the output. Due to the calculation adopted by the series thin-film evaporator in engineering was estimated process, there were large errors between the design output and the production when the parameters were determined. On the other hand, rough set theory is one of the math's tools to deal with the ambiguity and non-accuracy information. Any earlier knowledge besides the data aggregates is not needed. The uncertainty of the knowledge was expressed through the approximate between data, connotative knowledge was found through epagogic study, then the best decision can be chosen. At present, the study on applying the rough set theory to the chemical process parameter was just a new exploration. The paper first brought appliance of the rough set theory to the thin-film evaporator system forward. The significance of the study is as follows: The rough set theory was to the best advantage by applying it to deal with the process parameter of thin-film evaporator. The knowledge of forecasting rules was found by the training data. The production process was optimized and controlled. The difficulty of making certain the parameter was avoided by this method and the quantity of calculation was reduced. The reduct of the thin-film evaporator's information systems based on the rough set theory has reduced the size of the database. The studies of the importance of the parameters accorded with the results of the literature. The forecasting results and the test results of the output and evaporation intensity coincide well. The results of the Neural Network whose data was provided before and after the reduct are consistent. It was showed that applying the rough set theory to analysis the process parameters of the thin-film evaporator is feasible. The major research work and conclusions are summarized as follows: 1. The Rough set theory and its applications in the data mining and Knowledge Discovery in Database are summarized. This method is used to deal with the process data of the thin-film evaporator, then a new method to study the performances of thin film evaporator based on the data analysis is obtained. 2. The test flow and the data collection systems of the thin-film evaporator are introduced, different methods were used to deal with different data. The important condition factors to the output were analyzed and studied. Finally the information systems based on the rough set theory were created. 3. One of the data analysis saddlebags ROSETTA based on the rough set theory was adopted to realize the KDD course on the output forecasting of the thin-film evaporator. The forecasting result was gained and the rate of the forecasting achieved 71%. 4. One of the data analysis tools RSDAV based on the rough set theory was adopt to implement the analysis process based on the math's tool MATLAB. In the analysis process, the method, which combines the RSDAV and the neural network saddlebag, was advanced. So the rough neural network emluator was carried out. A satisfied correlation coefficient was received through the linearity regress of the result of the emluator and the goal output. 5. The way and the method to further study is put forward.