#TITLE_ALTERNATIVE#

• Main Author: PRANASARI (NIM : 23410051); Pembimbing : Dr. Ir. Rachmawaty Wangsaputra, TRISTANIA
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/17467
• Institution: Institut Teknologi Bandung

Chocolate candy is cocoa products made through several processes of mixing, refining, tempering, conching, and moulding. The quality of fineness of chocolate candy is expressed in particle size, particle density and organoleptic response. In refining process in the ball mill, there are three controllable factors, which are the diameter of the ball (A), the utility of the ball mill (B) and the duration of the refinement process (C), and also an uncontrollable factor operator skills (M). The determination of optimal setting in refining process can be set by experiment. Taguchi experimental design uses resources economically in producing robust products against uncontrollable factors and are able to analyze the factors that can produce the best response. Fundamental weaknesses in Taguchi method lays on restriction in handling multi-response problem and in the determination of the optimum setting at undesigned levels. The purpose of this study is to determine the optimal setting of chocolate candy production process, particularly in the refining process in the ball mill. This study also aims to develop methods of integrating the Taguchi method, Grey Relational Analysis (GRA) for handling multi-response problems, and methods of Backpropagation Network (BPN) to predict responses at undesigned levels. Research methodology includes a study of Taguchi method, GRA method especially the calculating of Grey Relational Grade (GRG) value using Signal to Noise Ratio (SNR) and the weighting using PCA method by calculating the eigenvector of each principal component of the response. Intensively, BPN method is studied. In the BPN method, two basic studies, BPN 1 and BPN 2, are conducted. BPN 1 maps each level of the multi-response quality factors and BPN 2 uses a single response approach, which divides the three responses to be processed individually in three BPN network. BPN 1 has a very large error, while BPN 2 can improve the performance of the BPN 1 but has weaknesses in its simplicity. Based on weaknesses encountered in BPN 1 and BPN 2, BPN 3 is developed. BPN 3 maps each level factor into a single response factor is expressed in the GRG. The results shows that BPN 3 gives smaller error and has a more simple structure than BPN 1 and BPN 2. From the integration of the Taguchi method, GRA and BPN, the obtained optimum settings in the process of refining is factor A at 30 mm, factor B at 200%, and factor C at 27 hours. The conclusion of this study is that the integration of the Taguchi method, GRA and BPN can be applied in the determination of the optimum setting of the refining process in the manufacture of chocolate candy and especially the use of BPN open possibility for determining the optimal setting eventhough the levels are not designed.