Sanjani, Vanisya Dewimarta (2025) Pengendalian Kualitas Proses Produksi Pupuk ZA III di PT Petrokima Gresik Menggunakan Diagram Kendali MEWMV dan MEWMA Berbasis Residual Model Multioutput Least Square Support Vector Regression (MLS-SVR). Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5003211066-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5003211066-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (7MB) | Request a copy |
Abstract
PT Petrokimia Gresik merupakan pabrik pupuk terlengkap di Indonesia yang memproduksi pupuk ZA III. Penggunaan pupuk ZA III membantu meningkatan kualitas dan kuantitas hasil panen. Oleh karena itu, kandungan dalam pupuk harus sesuai spesifikasi yang ditetapkan. Sehingga, diperlukan penerapan pengendalian kualitas secara statistik untuk memantau proses produksi pupuk ZA III. Karakteristik kualitas yang digunakan pada penelitian ini yaitu N (Nitrogen), H_2 O (Air), serta S (Sulfur). Ketiga karakteristik kualitas tersebut merupakan data multivariat yang saling berhubungan. Hubungan dari ketiga karakteristik kualitas menunjukkan adanya kasus autokorelasi. Adanya autokorelasi dapat menyebabkan false alarm yang membuat pengambilan keputusan menjadi tidak tepat. Untuk mengatasi kasus autokorelasi maka menggunakan model Multioutput Least Square Support Vector Regression (MLS-SVR). Input yang digunakan pada pemodelan MLS-SVR menggunakan lag dari model terbaik VAR yaitu VARI (5), sehingga pemodelan MLS-SVR dengan kernel RBF didapatkan hyper-parameter optimal yaitu γ^'=2^(-5),γ^''=2^(-8) dan σ= 2^(-3) dengan MSE sebesar 0,0045. Dari residual yang didapatkan dari pemodelan MLS-SVR akan digunakan untuk pembuatan diagram kendali Multivariate Exponentially Weighted Moving Variance (MEWMV) dan Multivariate Exponentially Weighted Moving Average (MEWMA). Pada diagram kendali MEWMV dengan pembobot ω=0,1, λ=0,3 dan L = 2,7949, variabilitas proses fase I telah terkendali secara statistik setelah dilakukan iterasi pengamatan out of control sebanyak 5 kali. Pada fase II, diketahui bahwa tidak terdapat pengamatan yang out of control sehingga variabilitas telah terkendali secara statistik. Selanjutnya pada diagram kendali MEWMA dengan nilai pembobot λ = 0,05, rata-rata proses pada fase I telah terkendali secara statistik setelah dilakukan iterasi pengamatan out of control sebanyak 8 kali. Pada fase II, diketahui tidak terdapat pengamatan yang out of control sehingga rata rata proses telah terkendali secara statistik. Pada analisis kapabilitas proses, secara univariat dan multivariat telah memenuhi batas spesifikasi yang telah ditetapkan sehingga kinerja proses produksi pupuk ZA III di PT Petrokimia Gresik telah kapabel.
=========================================================================================================================================
PT Petrokimia Gresik is the most complete fertilizer factory in Indonesia that produces ZA III fertilizer. The use of ZA III fertilizer helps increase the quality and quantity of crop yields. Therefore, the contents of the fertilizer must meet the specified specifications. Thus, the application of statistical quality control is needed to monitor the production process of ZA III fertilizer. The quality characteristics used in this study are N (Nitrogen), H_2 O (Water), and S (Sulfur). These three quality characteristics are multivariate data that are interrelated. The relationship among these three quality characteristics shows the presence of autocorrelation. The presence of autocorrelation can cause false alarms, leading to incorrect decision-making. To overcome the autocorrelation issue, the Multioutput Least Squares Support Vector Regression (MLS-SVR) model is used. The input used in the MLS-SVR model are derived from the lag of optimal VAR model, which is VARI (5). Based on the MLS-SVR modeling with an RBF kernel, the optimal hyper-parameters obtained are γ^'=2^(-5),γ^''=2^(-8) and σ= 2^(-3) , with a Mean Squared Error (MSE) of 0.0046. The residuals obtained from the MLS-SVR modeling will be used to construct the Multivariate Exponentially Weighted Moving Variance (MEWMV) and Multivariate Exponentially Weighted Moving Average (MEWMA) control charts. In the MEWMV control chart with weight value of ω=0,1, λ=0,3, and L = 2,7949, the process variability in Phase I was statistically in control after five iterations of out of control observations. In Phase II, there were no out of control observations, indicating that the process variability was statistically in control. Furthermore, in the MEWMA control chart with a weight value of λ=0,05, the process average in Phase I was statistically in control after eight iterations of out of control observations. In Phase II, there were no out of control observations, indicating that the process average was statistically in control. In the process capability analysis, the univariate and multivariate results show that the process has met the specified specification limits, indicating that the production process of ZA III fertilizer at PT Petrokimia Gresik is capable.
| Item Type: | Thesis (Other) |
|---|---|
| Uncontrolled Keywords: | MEWMA, MEWMV, MLS-SVR, Pengendalian Kualitas, Pupuk ZA III MEWMA, MEWMV, MLS-SVR, Quality Control, ZA III Fertilizer |
| Subjects: | H Social Sciences > HD Industries. Land use. Labor H Social Sciences > HD Industries. Land use. Labor > HD9980.5 Service industries--Quality control. |
| Divisions: | Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis |
| Depositing User: | Vanisya Dewimarta Sanjani |
| Date Deposited: | 31 Jul 2025 06:23 |
| Last Modified: | 31 Jul 2025 06:23 |
| URI: | http://repository.its.ac.id/id/eprint/124680 |
Actions (login required)
 |
View Item |