Soft Sensing Based on Improved GA-LSSVM

Some variables in industrial process are very difficult to measure on-line. To overcome this problem, a kind of soft sensing based on improved GA-LSSVM (IGA-LSSVM) is proposed in this research. First, KICA was used to extract main features of the data with high dimension patterns, and then an improved GA-LSSVM was established. This model not only utilizes the ability of quickly solving speed of LSSVM, but also the powerful global search performance of adaptive GA. Therefore the adaptability of LSSVM model is improved. The proposed method has been used for building soft sensing of diesel oil solidifying point. The result shows that IGA-LSSVM approach has high precision and good generalization ability.