In this paper

a soft sensor technique on the basis of support vector machines (SVM) was proposed to estimate the propylene concentration on the bottom of the distillation column

which took into consideration of the process knowledge and operation characteristics of the propylene distillation column.The advantage of SVM in small samples modeling was taken in the soft sensor technique.Furthermore

an intelligent control system was designed to real-time control the bottom propylene concentration in the distillation.A fuzzy PID was used as the master controller for the propylene concentration.One PID controller was used as the slave controller for the sensitive plate temperature.The two controllers constituted a cascade control system.In order to eliminate disturbance of the feed stock

a feedforward controller was designed in the system.The designed soft sensor and control strategy were used in one selected olefin plant and the results were demonstrated and discussed in this paper.The results were promising and showed that the constructed soft sensor was precise enough to satisfy the industrial on-site requirements.Moreover

the control strategy was efficient to guarantee the bottom propylene concentration under the specified limit.