Blast Furnace Expert System For The Exceptional Situation Forecast

With the rapid development of economics, China's steel production has been growing more than 20% since 4 years ago, and topped in the world for consecutive 9 years. In 2006, the energy consumption of the large and medium-sized Chinese steel enterprises was more than 300 million tons of standard coal, accounting the country's total energy consumption for about 15%, and environmental pollution has reached about 10%. Therefore saving energy and reducing pollution are the most important task for Chinese iron and steel enterprises. The blast furnace is the key point for iron and steel enterprises, and process of iron making is complex and changeable. This paper was made in the course of taking part in the project"Research and Application of BF Expert System for the Exceptional Situation Forecast", and the following works was completed simultaneously:⑤To unequal Spacing sequence made up with the monitor value of Titanium content, a general forecasting method combined with cubic spline and ARIMA model is proposed. First of all, equal spacing time-series was obtained from cubic spline function's interpolation on the original time-series. Next, ARIMA model of equal spacing time-series was made. Last, the future Titanium content was forecasted. The simulation analysis showed that its forecast accuracy is better, the error in shift point of Titanium curve is smaller, and this method improves the rate of success.⑥As for more parameters and more complex change, a scheme including two step treatments is proposed. The first stage of treatment is to get rid of noise and obtain valid data. The second step is to extract change feature of data deeply, such as larger fluctuations in air volume (Variance), Air volume dumping suddenly (Gradient), and air volume normal (Mean), for it could express the exceptional situation of BF accurately.⑦To reduce time of calculating and completing more computation of models, three-level cache program is proposed. In this program, the most data are saved in the memory as temporary storage space, the caches including the original data cache, the general data cache and the eigenvalue cache. Another, knowledge rules, threads hold value and other setting parameters are saved in the memory, to avoid getting them from database. The result of testing in scene indicates that the system can complete all model correctly, accounting time of CPU for about 10%.⑧On the basis of the completeness of design of expert system structure, Visual C++ and Microsoft SQL Server as development platforms, the expert system software code was completed. Over the course of development at the expert system, the ADO technology was used to achieve data access, and OPC client services was packaged into the system, which make it simple and easy to set OPC Service.