Rancang Bangun Sistem Pengukuran Viskositas Dan Klasifikasi Mutu Minyak Goreng Menggunakan Spektroskopi Near-Infrared (NIR) dan Artificial Neural Network (ANN)

Rahmadhani, Amalia (2026) Rancang Bangun Sistem Pengukuran Viskositas Dan Klasifikasi Mutu Minyak Goreng Menggunakan Spektroskopi Near-Infrared (NIR) dan Artificial Neural Network (ANN). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2042221072-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Penelitian ini bertujuan untuk merancang dan mengevaluasi sistem pengukuran serta prediksi viskositas minyak goreng berbasis spektroskopi Near-Infrared (NIR) yang terintegrasi dengan model Partial Least Squares (PLS) dan Artificial Neural Network (ANN). Sistem dikembangkan menggunakan sensor spektral AS7265x dan Raspberry Pi 5 sebagai unit pemrosesan, dengan data spektrum diproses menggunakan metode Standard Normal Variate (SNV), Savitzky–Golay smoothing, dan standard scaler. Validasi sensor dilakukan terhadap tiga jenis sampel larutan berwarna dengan sembilan variasi kondisi pengukuran. Berdasarkan analisis error, kondisi operasi optimal diperoleh pada kondisi 2 (25 mA arus LED dan jarak sensor 4 cm). Pada kondisi ini, sensor menghasilkan error rata-rata sebesar 3,43% dengan akurasi keseluruhan sebesar 96,57%, yang menunjukkan bahwa sensor memiliki tingkat kesesuaian yang baik terhadap alat referensi. Model PLS menunjukkan performa prediksi viskositas dengan nilai R² = 0,4008 dan RMSE = 4,5955 pada data pelatihan, serta R² = 0,3800 dan RMSEP = 4,9591 pada data pengujian, yang menunjukkan kemampuan generalisasi model yang cukup stabil. Nilai prediksi viskositas dan parameter warna digunakan sebagai masukan model ANN untuk klasifikasi mutu minyak goreng, yang menghasilkan akurasi sebesar 96% pada data pengujian. Evaluasi karakteristik statik alat ukur menunjukkan bahwa alat memiliki error rata-rata sebesar 5,50% dengan akurasi 94,50%, serta jangkauan pengukuran viskositas pada rentang 67,581–68,015 mPa·s dengan nilai span sebesar 0,434. Hasil penelitian ini menunjukkan bahwa meskipun sensor dan alat ukur memiliki akurasi yang baik, keterbatasan sensitivitas dan rentang pengukuran masih mempengaruhi kemampuan alat dalam merepresentasikan variasi viskositas minyak goreng secara optimal.
=================================================================================================================================
This study aims to design and evaluate a cooking oil viscosity measurement and prediction system based on Near-Infrared (NIR) spectroscopy integrated with Partial Least Squares (PLS) and Artificial Neural Network (ANN) models. The system was developed using an AS7265x spectral sensor and a Raspberry Pi 5 as a processing unit, with spectral data processed using the Standard Normal Variate (SNV), Savitzky–Golay smoothing, and standard scaler methods. Sensor validation was performed on three types of colored solution samples with nine variations of measurement conditions. Based on error analysis, the optimal operating conditions were obtained at condition 2 (25 mA LED current and 4 cm sensor distance). Under these conditions, the sensor produced an average error of 3.43% with an overall accuracy of 96.57%, indicating that the sensor has a good level of conformity to the reference tool. The PLS model showed good viscosity prediction performance with R² = 0.4008 and RMSE = 4.5955 on the training data, and R² = 0.3800 and RMSEP = 4.9591 on the testing data, indicating a fairly stable model generalization capability. The predicted viscosity values and color parameters were used as input to the ANN model for the classification of cooking oil quality, which resulted in an accuracy of 96% on the testing data. Evaluation of the static characteristics of the measuring instrument showed that the instrument had an average error of 5.50% with an accuracy of 94.50%, and a viscosity measurement range in the range of 67.581–68.015 mPa·s with a span value of 0.434. The results of this study indicate that although the sensor and measuring instrument have good accuracy, limitations in sensitivity and measurement range still affect the instrument's ability to optimally represent variations in cooking oil viscosity.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Minyak Goreng, Viskositas, Spektroskopi Near-Infrared, Artificial Neural Network, Klasifikasi Mutu. Cooking Oil, Viscosity, Near-Infrared Spectroscopy, Artificial Neural Network, Quality Classification.
Subjects:	Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QC Physics > QC100.5 Measuring instruments (General) Q Science > QC Physics > QC451 Spectroscopy T Technology > TA Engineering (General). Civil engineering (General) > TA1573 Detectors. Sensors
Divisions:	Faculty of Vocational > Instrumentation Engineering
Depositing User:	Amalia Rahmadhani
Date Deposited:	01 Aug 2026 06:18
Last Modified:	01 Aug 2026 06:18
URI:	http://repository.its.ac.id/id/eprint/141459

Actions (login required)

View Item