#TITLE_ALTERNATIVE#

• Main Author: SHOLIHUL FATHONI (NIM : 23315005), HAPPY
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/22367
• Institution: Institut Teknologi Bandung

Boilers are common industrial equipment that heat water to produce steam. The <br /> <br /> <br /> steam can be used for several purposes such as (1)for Steam Turbine Generator; <br /> <br /> <br /> and (2)Heating Process. In a boiler, a steam drum has a function to contain water <br /> <br /> <br /> that will be heated and then generate steam. The steam requirement for the process <br /> <br /> <br /> can be vary, so the right control methods are needed to maintain the water level in <br /> <br /> <br /> the steam drum appropriately. If the water is too low, the overheating can be <br /> <br /> <br /> detrimental to the steam drum, on the contrary, the higher of the water level <br /> <br /> <br /> compared to the set point can lead to the ineffective steam production. <br /> <br /> <br /> The problem in this research is how to control the steam drum water level <br /> <br /> <br /> appropriately, despite the incoming disturbances. Disturbance to water level occur <br /> <br /> <br /> due to the variation of the outlet valve opening to adjust the steam needed in the <br /> <br /> <br /> process. The usual method to control the level in the steam drum is PID. The PID <br /> <br /> <br /> method is one of the oldest methods that is still effective to control the water level. <br /> <br /> <br /> The purpose of this study is to classify the inlet valve opening based on the training <br /> <br /> <br /> data to minimize the effect of disturbances on a steam drum boiler by using Support <br /> <br /> <br /> Vector Machine (SVM) method. By minimizing the influence of the disturbances, <br /> <br /> <br /> the water level can be controlled based on a set point. Step to use SVM method are: <br /> <br /> <br /> (1) Obtain the training data set; (2)Training based on the obtained data; (3)Testing <br /> <br /> <br /> the generated model; (4)Conducting test on measured data; (5)comparing the results <br /> <br /> <br /> of the measured data with the output data of SVM. <br /> <br /> <br /> The classification percentage of Manipulated Variable (MV) using SVM method <br /> <br /> <br /> has shown agreeable error when comparing with the measurement results. The <br /> <br /> <br /> detail can be described as follows: (1)Training data 1 with the switch off and on condition of the inlet pump has Root Mean Square Error (RMSE) of 3.3%; <br /> <br /> <br /> (2)Training Data 2 has a RMSE of 1.6% while the pass area of the outlet valve as <br /> <br /> <br /> disturbance were changed between 25%-100% with 5% step; (3)Training data 3 has <br /> <br /> <br /> RMSE about 2% while the pass area of the outlet valve as disturbance were changed <br /> <br /> <br /> between 30%-100% with 10% step; (4)Training data 4 has RMSE of 2.3% while <br /> <br /> <br /> the pass area of the outlet valve as disturbance were changed between 100%-30% <br /> <br /> <br /> with 10% step; (5)Training data 5 has RMSE of 2.7% while changing the set point. <br /> <br /> <br /> It can be concluded that by using the SVM method, the water level in the steam <br /> <br /> <br /> drum boiler can be controlled properly because the well classification of the inlet <br /> <br /> <br /> valve opening. The RMSE values that occurred in the five experimental categories <br /> <br /> <br /> were 3.3%; 1.6%; 2%; 2.3%; and 3.5%. In other words, this SVM method can be <br /> <br /> <br /> an alternative controller to maintain the level of steam drum in addition to PID.