Optimasi Berbasis AI Terhadap Kinetika Fermentasi Waktu Singkat Pada Yogurt Berbasis Kedelai Menggunakan Data Sensor Real-Time.

Hapnifaturrosidah, ST. Hapnifaturrosidah and Safitri, Siti Nurlaia Ayu (2026) Optimasi Berbasis AI Terhadap Kinetika Fermentasi Waktu Singkat Pada Yogurt Berbasis Kedelai Menggunakan Data Sensor Real-Time. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2041221127-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Produksi yogurt berbasis kedelai umumnya memerlukan waktu fermentasi lebih lama dibandingkan yogurt susu hewani karena tingginya kapasitas penyangga protein kedelai. Penelitian ini bertujuan untuk mengetahui pengaruh parameter fermentasi terhadap kinetika fermentasi waktu singkat (Short-Time Fermentation), mengembangkan dan mengevaluasi model machine learning berbasis algoritma Random Forest untuk memprediksi perubahan pH, serta menentukan waktu fermentasi optimum berdasarkan hasil prediksi model. Fermentasi dilakukan menggunakan soy milk powder dengan variasi konsentrasi 8%, 10%, dan 12%. Perubahan pH dan suhu dipantau menggunakan sistem sensor, sedangkan kinetika fermentasi dianalisis menggunakan model Gompertz. Data waktu fermentasi, konsentrasi soy milk powder, volume fermentasi, suhu, dan kondisi pengadukan digunakan sebagai variabel masukan Random Forest dengan pH sebagai variabel target. Hasil penelitian menunjukkan bahwa waktu fermentasi berpengaruh terhadap kinetika fermentasi yang ditandai dengan penurunan pH secara bertahap akibat meningkatnya aktivitas metabolisme bakteri asam laktat selama fermentasi. Selain itu, variasi konsentrasi soy milk powder memengaruhi kinetika fermentasi secara nonlinier. Konsentrasi 8% menghasilkan fermentasi tercepat dengan nilai μmax tertinggi sebesar 0,9524 jam⁻¹ dan λ terendah sebesar 3,9876 jam, sedangkan konsentrasi 10% menghasilkan μmax terendah sebesar 0,7258 jam⁻¹ dan λ tertinggi sebesar 4,3573 jam. Model Gompertz menunjukkan kesesuaian yang sangat baik terhadap data eksperimen dengan nilai R² > 0,98. Model Random Forest menghasilkan MSE sebesar 0,2192, RMSE sebesar 0,4682, MAE sebesar 0,4328, dan R² sebesar 0,7420, yang menunjukkan kemampuan model dalam mengikuti tren perubahan pH selama fermentasi. Berdasarkan prediksi model pada proses scale-up dengan volume fermentasi 1500 mL, waktu fermentasi optimum untuk mencapai pH 4,5 diperkirakan sebesar 13,8 jam, dan hasil validasi menunjukkan model mampu mengikuti tren perubahan pH pada volume fermentasi yang lebih besar. Dengan demikian, integrasi model Gompertz dan Random Forest berpotensi digunakan untuk menganalisis kinetika fermentasi, memprediksi perubahan pH, serta menentukan estimasi waktu fermentasi optimum pada produksi yogurt berbasis kedelai. ===================================================================================================================================
Soy-based yogurt production generally requires a longer fermentation time than dairy yogurt due to the high buffering capacity of soy proteins. This study aimed to investigate the effects of fermentation parameters on Short-Time Fermentation kinetics, develop and evaluate a Random Forest-based machine learning model to predict pH changes, and determine the optimum fermentation time based on the model predictions. Fermentation was carried out using soy milk powder at concentrations of 8%, 10%, and 12%. Changes in pH and temperature were monitored using a sensor-based system, while fermentation kinetics were analyzed using the Gompertz model. Fermentation time, soy milk powder concentration, fermentation volume, temperature, and stirring conditions were used as input variables for the Random Forest model, with pH as the target variable. The results showed that fermentation time significantly affected fermentation kinetics, as indicated by the gradual decrease in pH resulting from the increased metabolic activity of lactic acid bacteria during fermentation. In addition, variations in soy milk powder concentration influenced fermentation kinetics in a nonlinear manner. The 8% concentration resulted in the fastest fermentation, with the highest μmax value of 0.9524 h⁻¹ and the lowest λ value of 3.9876 h, whereas the 10% concentration produced the lowest μmax value of 0.7258 h⁻¹ and the highest λ value of 4.3573 h. The Gompertz model showed excellent agreement with the experimental data, with an R² value greater than 0.98. The Random Forest model achieved an MSE of 0.2192, RMSE of 0.4682, MAE of 0.4328, and R² of 0.7420, indicating its capability to capture the trend of pH changes during fermentation. Based on the model predictions for the scale-up process at a fermentation volume of 1500 mL, the optimum fermentation time to reach pH 4.5 was estimated to be 13.8 h, and the validation results demonstrated that the model was able to follow the pH trend at a larger fermentation volume. Therefore, the integration of the Gompertz and Random Forest models has the potential to analyze fermentation kinetics, predict pH changes, and estimate the optimum fermentation time for soy-based yogurt production.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Short-Time Fermentation, Artificial Intelligence, Machine Learning
Subjects:	T Technology > TP Chemical technology T Technology > TP Chemical technology > TP156 Crystallization. Extraction (Chemistry). Fermentation. Distillation. Emulsions.
Divisions:	Faculty of Vocational > 24305-Industrial Chemical Engineering Technology
Depositing User:	S.T. Hapnifaturrosidah Hapnifaturrosidah
Date Deposited:	04 Aug 2026 01:00
Last Modified:	04 Aug 2026 01:00
URI:	http://repository.its.ac.id/id/eprint/142873

Actions (login required)

View Item