建構以遺傳演算法結合塑模工具為基礎之玻璃工廠模擬排程系統

Abstract

資訊科技的發展，帶動了企業部門資訊化，為了達成規模經濟的目標，企業無不在生產流程、物料管理、交易處理資料上擴充新技術的資訊系統，以求有效率、結構化的處理各部門業務。供應鏈管理﹙Supply Chain Management, SCM﹚與企業資源規劃﹙Enterprise Resource Planning, ERP﹚系統的導入與實作，則是目前各企業致力於提昇整體競爭力、與強化上下游產業體系多層次分工的解決方案。傳統人工排程系統受限於主事者對於生產線的狀況掌握，諸如原料供應、機器產能、生產程序等，在產品種類不多，與生產線產品單一化的情況下或許表現還能令人滿意，一但超出了這個範圍，繁雜的排程問題與最佳化的追求目標將是人力所不及的領域。 本研究藉由整合不同的分析理論，運用斐氏網路﹙Petri Net﹚將生產線模型化，並進行工廠加工機具的狀態配置、運作流程定義，配合彈性製造系統﹙FMS﹚對於生產線自動化的定義，結合遺傳演算法﹙Genetic Algorithm﹚快速狀態空間求解的特性，進行排程分析與模擬，以便在有限時間內尋找最佳與替代的排程策略。研究中實際建構應用在製造業玻璃加工廠，具備GA與PN特性的即時排程模擬系統﹙Genetic Algorithm – Petri Net Scheduling Simulation System, G-PSS﹚；透過時間導向/工作數量導向的參數設定，期望在同時能夠生產多種不同產品的原則下，藉由電腦快速的模擬與評估排程計劃，最後將計劃表送至工廠生產線上執行。 本研究主要貢獻為改善傳統人工排程耗時耗力，又不能滿足單一生產線多樣產出﹙少量多樣﹚的問題、減少人員對生產線流程掌控能力的高度要求、減少執行工作排程所需的專業知識背景、增加生產線產能、減少工作機在使用率及閒置時間上的浪費、縮短訂貨至交貨的間隔時間，達成規模經濟下生產排程最佳化，與成本控制、生產資源配置最適化的目標。

As the development of IT that brought industry into information age, industry has tried by all means to expend the new IT system of manufacturing process, material management, and trading data to deal with business with more efficient and organizational methods as to achieve the goal of model business. The adoption and practice of Supply Chain Management (SCM) and Enterprise Resource Planning (ERP) are just the solution to raise competitive power and reinforce the multiple-labor-division of industry. Traditional man-made scheduling system is limited with the manager’s control of the production line, such as material supply, machine property, manufacturing process…etc. The result could be content if the product variety is limited and the production line is focused on merely one product. However, once goes out of the range, the complex scheduling and the pursuit of optimization will go beyond the human ability. With the convergence of different analysis theory, we adopt Petri Net to model the production line and take the transformation system of the Petri Net to depict the system. In cooperating with the definition of production line automation given by FMS, we combine it with the characteristic of GA in solving the problem of swelling constitution to progress the analysis and simulation of production schedule. Drawing on the process mentioned, we seek for the optimal and minor optimal scheduling rules to modify the time-consuming man-made scheduling and laborious on production line. In this study, we developing a Genetic Algorithm – Petri Net Scheduling Simulation System (G-PSS) with real case of glass operate factory in manufacturing industry. We can setup the Time Oriented/Job Oriented parameter, and the scheduling system would process the analysis and simulation for manufacturing different products in one single production line. After simulation, the scheduling plan would be carried out in the production line. This study contributes most in developing a production scheduling system that combines Petri Net and GA characteristics for solving the problem of time consuming, laborious, and can’t produce various products in a single production line of tradition al man-made scheduling system. Besides the optimal scheduling strategy, it provides the simulation device on line to amend and evaluate each scheduling more efficiently and improve the production and competition ability of the industry. For the prospect, we are going to study further on the application of G-PSS system in the ERP production scheduling model system.