Soft sensor based on least square support vector machine with limited memory for crystal particle size prediction in PTA purification process

The accurate estimation of the average crystal particle size in PTA purification process is of fundamental importance in process monitoring, advanced control and optimization. A method of least square support vector machine with limited memory is applied to model soft senor to predict the average particle size on-line. Compared with traditional least square support vector machine, Small samples in fixed length are used to train a drifting model in small scale which need not optimize its structure by pruning algorithm. Some useful information would be lost if old samples are discarded directly so an idea is proposed to introduce the information to the model by the maximum values and minimum values of overall samples... The results of simulation show that the soft sensor based on the proposed method has high precision and is suitable for time-varying system with samples which distribution is not uniform.