PCR-TOPSIS MULTIRESPONSE OPTIMIZATION BASED-ON TAGUCHI AND NEURAL NETWORK APPROACHES FOR WASTEWATER TREATMENT
Devi Putri Isnarwaty (a*), Muhammad Mashurib(b), Hidayatul Khusna (b)

a) Department of Statistic, Faculty of Science and Data Analytics, Institut Teknologi Sepuluh Nopember, Indonesia
*devisnarwaty[at]its.ac.id
b) Department of Statistic, Faculty of Science and Data Analytics, Institut Teknologi Sepuluh Nopember, Indonesia

Abstract

Waste is the result of waste from a production process, both industrial and domestic (household). Water treatment is carried out by adding several treatments such as adding coagulants or adjusting the stirring setting. This is done in order to normalize pH levels and reduce Total Solid Suspended (TSS) levels in wastewater, so it is necessary to optimize the level of treatment. This optimal level of administration is carried out in order to minimize the cost of wastewater treatment and achieve targets according to predetermined quality standards. In achieving an effective waste treatment process and maintaining good water quality, the researchers intend to optimize the multiresponse Process Capabilty Ratio-Technique for Order Preference by Similarity to Ideal Solution (PCR-TOPSIS) with the Taguchi design using a neural network approach. The Taguchi method is able to evaluate several factors with a minimum number of tests, while the Neural Network can be used to predict PCR-TOPSIS if the combination of factor levels that optimizes the response is not in the Orthogonal Array (OA) design. The utilized method can define that the optimal factor level combination of the waste treatment process. The best model combination criteria are selected based on the value of R2 and Mean Square Error (MSE). In this study, the OA design used is mixed level L8 which means that there is 1 factor that has 4 levels and 2 factors that have 2 levels. The results of the ANOVA calculations show that the optimum conditions can be achieved at the combination of the A4B1C2 level, namely Aluminum Sulfate (Al2(SO4)3) of 50 mg/L, stirring speed of 200 rpm and stirring time of 1 minute. The prediction results of the PCR-TOPSIS Neural Network were not significantly different from the prediction results of the multiresponse Taguchi PCR-TOPSIS method

Keywords: Neural Network, Orthogonal Array, PCR-TOPSIS, Taguchi, Wastewater Treatment

Topic: MATHEMATICS AND STATISTICS