Pengendalian Proses Produksi Kaca Gelap Menggunakan Diagram Kendali Multivariat T2 Berbasis Spatial Sign Covariance Matrix

Teduhati, Grahsaro Yosha (2026) Pengendalian Proses Produksi Kaca Gelap Menggunakan Diagram Kendali Multivariat T2 Berbasis Spatial Sign Covariance Matrix. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5003221015-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Cacat top haze pada produksi kaca gelap di PT. Asahimas Flat Glass, Tbk. dipengaruhi oleh empat karakteristik kualitas yang saling berkorelasi, yaitu komposisi pewarna, suhu, atmosfer, dan waktu tinggal kaca. Pengendalian kualitas pasca-produksi yang selama ini diterapkan dinilai kurang efektif dalam mendeteksi ketidakstabilan proses secara dini, sementara data proses produksi umumnya mengandung autokorelasi, melanggar asumsi normalitas multivariat, dan berpotensi mengandung pencilan, sehingga diagram kendali T²-Hotelling klasik menjadi kurang sensitif dalam mendeteksi proses yang tidak terkendali. Penelitian ini bertujuan untuk mengendalikan proses produksi kaca gelap menggunakan diagram kendali multivariat T² berbasis Spatial Sign Covariance Matrix (SSCM) serta mengidentifikasi kapabilitas prosesnya. Sebelum pembentukan diagram kendali, dilakukan pemodelan MLS-SVR berdasar pada orde VAR untuk mengatasi autokorelasi pada data, kemudian SSCM digunakan sebagai estimator robust berbasis arah data terhadap residual MLS-SVR untuk mengatasi pelanggaran asumsi normalitas dan keberadaan pencilan, dengan Batas Kendali Atas (UCL) ditentukan melalui metode trimmed bootstrapping. Kapabilitas proses diukur menggunakan indeks kapabilitas non-normal multivariat (MCNpk). Hasil analisis menunjukkan bahwa keempat karakteristik bersifat dependen, residual model MLS-SVR tidak memiliki autokorelasi namun tidak berdistribusi normal multivariat. Diagram kendali T² berbasis SSCM pada Fase I berhasil mencapai kondisi in-control dengan UCL sebesar 0,488 setelah penanganan pencilan, sementara pada Fase II monitoring masih ditemukan observasi yang berada di luar batas kendali, menunjukkan bahwa proses produksi belum sepenuhnya stabil. Identifikasi lebih lanjut menunjukkan bahwa kombinasi karakteristik atmosfer dan waktu tinggal kaca menjadi kontributor utama ketidakstabilan proses. Nilai indeks kapabilitas proses yang diperoleh sebesar MCNpk = 0,918, mengindikasikan bahwa proses produksi belum kapabel secara multivariat dalam memenuhi spesifikasi standar mutu perusahaan. Dapat disimpulkan bahwa pendekatan diagram kendali T² berbasis SSCM efektif untuk memonitor proses produksi kaca gelap yang tidak berdistribusi normal multivariat, namun diperlukan peningkatan proses lebih lanjut untuk meningkatkan kapabilitas produksi terkait cacat top haze.
=====================================================================================================================================
The top haze defect in dark glass production at PT. Asahimas Flat Glass, Tbk. is influenced by four correlated quality characteristics: colorant composition, temperature, atmosphere, and glass residence time. The post-production quality control currently implemented is considered less effective in detecting process instability at an early stage, while production process data generally contain autocorrelation, violate the multivariate normality assumption, and potentially contain outliers, rendering the classical T²-Hotelling control chart less sensitive in detecting out-of-control processes. This study aims to control the dark glass production process using a Spatial Sign Covariance Matrix (SSCM)-based multivariate T² control chart and to identify its process capability. Prior to constructing the control chart, MLS-SVR modeling based on VAR order was performed to address the autocorrelation in the data, after which SSCM was employed as a data-direction-based robust estimator on the VAR residuals to address violations of the normality assumption and the presence of outliers, with the Upper Control Limit (UCL) determined using the trimmed bootstrapping method. Process capability was measured using the multivariate non-normal capability index (MCNpk). The analysis results indicate that the four characteristics are dependent on one another, and the residuals of the MLS-SVR model exhibit no autocorrelation but do not follow a multivariate normal distribution. The SSCM-based T² control chart in Phase I successfully achieved an in-control condition with a UCL of 0.488 after outlier treatment, while monitoring in Phase II still identified observations falling outside the control limits, indicating that the production process is not yet fully stable. Further identification revealed that the combination of the atmosphere and glass residence time characteristics is the main contributor to process instability. The obtained process capability index value is MCNpk = 0.918, indicating that the production process is not yet capable, in a multivariate sense, of meeting the company's standard quality specifications. In conclusion, the SSCM-based T² control chart approach is effective for monitoring a dark glass production process that does not follow a multivariate normal distribution; however, further process improvement is required to enhance production capability with respect to the top haze defect.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Diagram Kendali T2, Kapabilitas Proses, Kaca Gelap, Spatial Sign Covariance Matrix, Top Haze, T2 Control Chart, Capability Process, Dark Glass, Spatial Sign Covariance Matrix, Top Haze
Subjects:	H Social Sciences > HA Statistics 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:	Grahsaro Yosha Teduhati
Date Deposited:	03 Aug 2026 03:55
Last Modified:	03 Aug 2026 03:55
URI:	http://repository.its.ac.id/id/eprint/139848

Actions (login required)

View Item