Development of Soft Sensor with Time Difference of Process Variables Approach

• Main Author: Balakrishnan, Kala Krissna
• Format: Final Year Project
• Language: English
• Published: Universiti Teknologi PETRONAS 2012
• Subjects: TP Chemical technology
• Online Access: http://utpedia.utp.edu.my/9706/1/2012%20-%20Developent%20of%20Soft%20Sensors%20with%20Tie%20Differece%20of%20Process%20Variables%20Approach.pdf; http://utpedia.utp.edu.my/9706/

Soft sensors are used to estimate the process variables that are hard to measure online in a process unit but the predictive accuracy of the estimation will deteriorate due to certain reasons. The reasons are usually due to the changes of plant state, catalyst performance loss, sensor or process drift and scale deposition. In order to overcome the degradation of the soft sensors due to process drift, time difference of process variables is proposed to use for the predictive model. The objective of this paper is to develop data-driven soft sensors with time difference of process variables and to evaluate its advantages over traditional static soft sensors. The modeling technique used for this approach is Partial Least Squares (PLS) method. Partial least squares method is a numerical method based on multiple regression. The main purpose of PLS is to predict a set of dependent variables from a set of independent variables or predictors. In this paper, a binary distillation column is selected as a case study and its virtual plant is built in Hysys enviromnent. In the simulation, the input variables such as feed temperature, reflux flow rate, feed flow rate and steam flow rate are varied and the output data are captured with time. In addition, different sets of data were formed with various time differences in the variables. Those data are used to develop the soft sensor model using PLS technique in SIMCA-P software. The performance of the model is evaluated and compared with the conventional soft sensor. Based on the results, the predictive ability of the developed model is higher than the static conventional model.