Understanding the properties of starch, which is the main component of the most abundant food consumed by people, may benefit in creating desirable properties in the starch-based food we consume. Many research has been conducted in formulating starch mixtures to overcome undesirable effects such as having less water binding capacity, and poor mouthfeel due to imbalanced texture resulting from the starch processing stages. However, the effect of pH seems to have not received any significant interest. The fact that different food might be processed at different pH initiated the curiosity to identify the effect of pH in three different starches; normal maize, waxy maize, and potato, and the effect of pH on the starch properties after combined with different composition of soy protein (10-50%) and guar gum (1-3%). Pasting and gelation properties such as peak viscosity, holding strength, breakdown, setback, final viscosity, peak temperature, and gel strength were observed using a Rapid Visco Analyzer (RVA) and texture analyser, respectively. In addition to that, to support our findings, removal of surface protein and lipid from normal maize starch using Sodium dodecyl sulphate (SDS) was performed. The experiments showed a significant increase in the peak viscosity and decrease in peak temperature when the pH is increased in normal maize which contained a higher protein and lipid content than the other starches. Both normal maize and potato starch mixed with soy showed the highest decrease in peak viscosity and breakdown at pH 8 compared to the pH 4 and pH 7. On the other hand, changing the pH of the solution did not have a significant effect on all guar gum and starch mixtures. Lastly our results suggested that the strength of gel decreases with the increase in protein solubility.
Di antara sebagian besar sektor industri lainnya, industri kimia sedang mengalami pergolakan signifikan yang didorong oleh konsep yang secara kolektif dikenal sebagai Industri 4.0. Data sains adalah komponen penting dari Industri 4.0 karena memungkinkan ekstraksi informasi kontekstual dari berbagai sumber data. Ketika sistem menjadi lebih kompleks, kebutuhan para insinyur untuk mengekstrak sinyal dari data dengan tepat berkembang secara dramatis, menuntut literasi data dan keahlian analitik pada generasi berikutnya dari lulusan teknik kimia. Salah satu dari banyak kasus di mana data sains dan machine learning dapat diterapkan adalah untuk prediksi. Prediksi berbasis machine learning dapat diterapkan pada banyak aspek teknik kimia contohnya pada Chemical Engineering Plant Cost Index (CEPCI). CEPCI sangat penting untuk perhitungan desain pabrik dan dipengaruhi oleh banyak variabel. Pendekatan machine learning diperlukan untuk memperhitungkan semua variabel tersebut dan mendapatkan hasil yang tepat untuk variabel yang ditargetkan. Dengan demikian, tujuan dari tugas akhir ini adalah merancang program yang mampu memprediksi CEPCI. Alhasil, model regresi yang telah dibuat mampu memprediksi Composite CE Index dengan error rata-rata 3.75% dari index aslinya. ......Among most other industrial sectors, the chemical industry is undergoing a significant upheaval driven by concepts known collectively as Industry 4.0. Data science is an important component of Industry 4.0 since it enables the extraction of contextualized information from a variety of data sources. As systems become more complex, the necessity for engineers to appropriately extract signal from data develops dramatically, demanding data literacy and analytics expertise in the next generation of chemical engineering graduates. One of the many cases where data science and machine learning can be applied to is for prediction. Machine Learning based prediction can be applied to many chemical engineering aspects, in this case the Chemical Engineering Plant Cost Index (CEPCI). CEPCI is essential for plant design calculations and is greatly affected by numerous variables. Machine learning approach is needed to account for all said variables and obtain valid result for target variables. Thus, the purpose of this thesis is to design programs that are able to predict CEPCI. As a result, the regression model created was able to predict the Composite CE Index with average error of 3.75% from the real index.