#TITLE_ALTERNATIVE#

• Main Author: , Suyanto
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/9420
• Institution: Institut Teknologi Bandung

ABSTRACT: <br /> <br /> <br /> The data of infrared spectrophotometer is analyzed based on the Beer Law. The Beer Law states that the absorption value (A) of a sample test depends on the molar absorptivity (c), the cell path (b) and the sample concentration (C). The absorption value can be determined from measurement data by quantitative analysis of multicomponent. The result of the quantitative analysis plays a role as an approach of a single-channel state-space variable model of infrared spectrophotometer data, determined by Minimum-Covariance Deconvolution Method (MCD) which is developed based on the Mendel Theorem. The process noise (with variance Q) and the measurement noise (with variance R) have an effect on observation error of the absorption value of each output component of the model. For that reason, the application of Kalman filter with MCD method is needed to reduce the noise level generated by the process itself and data measurement of infrared spectrophotometer. The Kalman filter was applied to a single channel state-variable model obtained by MCD method from the measurement data of the absorption value of 8th component. The simulation was done to obtain the observation error, the relative error on absorption value and the relative error on component concentration either in steady state condition or transient condition. By introducing the process noise variance Q=1 and the measurement noise variance R =10, the application of Kalman filter succesfully reduce the noise level, so that the observation error, the relative error on absorption value and the relative error on component concentration were also reduced. The result of simulation, obtained by application of Kalman filter, is compared to the result that was calculated by Beer Law and the other that the Kalman filter was not applied. Those results is compared as follow: <br /> <br /> <br /> a) The calculation result obtained by the Beer Law gives an observation error equal to 0.01, relative error on concentration component equal to 2.72 % and relative error on absorption value equal to 0.4340. These values are used as a reference of other methods. <br /> <br /> <br /> b) In case Kalman filter was not applied (non-filtered results), the simulation gives an observation error equal to 0.0069, relative error on concentration component equal to 1.8768 % and relative error on absorption value equal to 0.2995. <br /> <br /> <br /> c) Application of Kalman filter gives the following results (filter results): 1) For steady state condition, the simulation gives an observation error equal to 0.0104, relative error on concentration component equal to 2.8288 % and relative error on absorption value equal to 0.4515. 2) For transient condition, the simulation gives an observation error equal to 0.0104, relative error on concentration component equal to 2.8288 % and relative error on absorption value equal to 0.4515. <br /> <br /> <br /> The Minimum-Covariance Deconvolution method (MCD), which was used in this research, gives the best result compared to the other methods. For example, by introducing the same values of the noise variances (Q=1, R=10), VAN CITTERT method gives covariance error equal to 0.7693. KALMAN method gives covariance error equal to 0.3602. JANSSON method gives covariance error equal to 0.7693. GOLD method gives covariance error equal to 0.3133 and MCD gives covariance error equal to 0.0104.