Wiratno, Stefanus Eko (2026) Pengembangan Model Holding Time Lintasan Produksi Tanpa Buffer dengan Distribusi Phase-Type dan Matrix-Exponential Guna Memperoleh Solusi Matriks-Analitik Eksak untuk Fungsi Distribusi Kinerja. Doctoral thesis, Institut Teknologi Sepuluh Nopember.
|
Text
02411860012003-Doctoral.pdf - Accepted Version Restricted to Repository staff only Download (7MB) | Request a copy |
Abstract
Lintasan produksi tanpa buffer umumnya dianalisis dengan model Markov atau dengan metode aproksimasi. Model Markov mengalami ledakan ruang keadaan ketika jumlah stasiun atau orde distribusi bertambah, sedangkan metode aproksimasi hanya memberikan hasil perkiraan dan terbatas pada nilai rata-rata. Penelitian ini mengembangkan model evaluatif yang eksak dan bekerja pada Tingkat fungsional, yaitu menghasilkan fungsi distribusi kumulatif (CDF) dari variabelvariabel lintasan produksi, bukan hanya nilai rata-ratanya. Model dasar yang digunakan adalah model holding time (HTM) dari Muth (1984), yang dikembangkan dengan memanfaatkan distribusi phase-type (PH), distribusi matrixexponential (ME), operasi Kronecker, serta sifat ketertutupan kedua kelas distribusi tersebut. Penelitian ini menghasilkan tiga model. Model 1 menurunkan rumus bentuk tertutup fungsional untuk lintasan produksi reliable tanpa buffer dengan tiga stasiun, termasuk waktu antar permintaan bahan baku dan waktu antar keberangkatan produk jadi, dan dapat menerima waktu layanan stasiun tengah yang berdistribusi bebas. Model 2 mengembangkan solusi matriks-analitik eksak untuk lintasan produksi dengan jumlah stasiun sembarang dan waktu layanan PH yang independen dan tidak identik, beserta algoritma penyelesaiannya. Model 3 memperluas Model 2 dari representasi PH ke representasi ME melalui transformasi similaritas matriks sehingga waktu komputasi berkurang secara signifikan. Validasi terhadap model Markov dan simulasi menunjukkan kesesuaian yang tinggi: selisih relatif terbesar throughput rate terhadap model Markov adalah 0,168%, dan waktu komputasi untuk lintasan produksi 13 stasiun sekitar 1,3 detik. Model 3 mempercepat komputasi sekitar 670 hingga 4.200 kali dibandingkan representasi PH asli. Model yang dihasilkan juga memungkinkan analisis sensitivitas dan implikasi manajerial, misalnya pengaruh variabilitas waktu proses dan penempatan stasiun terhadap kinerja lintasan produksi.
========================================================================================================================================
Production lines without buffers and with non-exponential processing times are commonly analyzed with Markov models or approximation methods. Markov models suffer from state-space explosion as the number of stations or the order of the distributions grows, while approximation methods only give estimates and are usually limited to mean values. This dissertation develops an exact evaluative model that works at the functional level, that is, it yields the cumulative distribution functions (CDFs) of the production line variables rather than only their means. The base model is the holding time model (HTM) of Muth (1984), extended by exploiting phase-type (PH) distributions, matrix-exponential (ME) distributions, the Kronecker operations, and the closure properties of both distribution classes. Three models are developed. Model 1 derives functional closed-form formulas for reliable three-station production lines without buffers, including the interdemand and interdeparture times, and allows the service time of the middle station to follow an arbitrary distribution. Model 2 develops an exact matrix-analytic solution for lines with an arbitrary number of stations and independent, non-identical PH service times, together with a solution algorithm. Model 3 extends Model 2 from PH to ME representations through matrix similarity transformations, which reduces the computation time substantially. Validation against Markov models and simulation shows close agreement: the largest relative difference in throughput rate against the Markov model is 0.168%, and the computation time for a 13-station line is about 1.3 seconds. Model 3 speeds up the computation by roughly 670 to 4,200 times relative to the original PH representation. The resulting models also support sensitivity analysis and managerial implications, such as the effect of processingtime variability and station placement on production line performance.
| Item Type: | Thesis (Doctoral) |
|---|---|
| Uncontrolled Keywords: | lintasan produksi, model holding time, distribusi phase-type, distribusi matrixexponential, solusi matriks-analitik, fungsi distribusi kumulatif, production line, holding time model, phase-type distribution, matrixexponential distribution, matrix-analytic solution, cumulative distribution function |
| Subjects: | Q Science > QA Mathematics > QA273.6 Weibull distribution. Logistic distribution. Q Science > QA Mathematics > QA274.7 Markov processes--Mathematical models. |
| Divisions: | Faculty of Industrial Technology and Systems Engineering (INDSYS) > Industrial Engineering > 26001-(S3) PhD Thesis |
| Depositing User: | Stefanus Eko Wiratno |
| Date Deposited: | 17 Nov 2026 08:21 |
| Last Modified: | 17 Nov 2026 08:21 |
| URI: | http://repository.its.ac.id/id/eprint/146068 |
Actions (login required)
![]() |
View Item |
