Visualization Assisted Decision Support for Capacity Planning

• Main Authors: Wen-Lin Kuo, 郭玟琳
• Other Authors: Tien-Lung Sun
• Format: Others
• Language: zh-TW
• Published: 1999
• Online Access: http://ndltd.ncl.edu.tw/handle/31555382124495778001

碩士 === 元智大學 === 工業工程研究所 === 87 === Production planning in a dynamic environment involves the evaluation of many ever changing, uncertain, and mutually interacting variables such as cost, product life cycle, manufacturing technology, market trend, enterprise resource allocation strategy, politics and many others. It is very difficult to build a computational model to automatically evaluate these dynamic variables and select the best plans. Current practice still requires human being to be involved in the decision-making process. To assist human planner comprehend the data, visualization of graphical display of the data is often employed. Visualization of data takes advantages of people’s natural strength in rapid visual pattern recognition and enables human planner to quickly and easily examine and comprehend different plans. In this work, 2D grid graphs are employed to visualize tabular data of production plans and their corresponding idle cost. In this content, it is often desired to assign adjacent points in the graph whose values are close the same color so that production patterns such as “the production amount of product i forms two clusters during the planning period” and “from month a to b, the production amount of products i to j are close” could be easily visualized. To do so, we assign each point in the graph a weight based on its adjacent points that have the same color. The summation of weights of points in the graph whose color is C is then defined as the “weighted area” of C. By finding a color-value mapping that assigns each color near-uniform weighted area through a combinatorial optimization using simulation annealing, the resulted color-value mapping would assign adjacent points in the graph whose values are close the same color. Another problem we addressed in this work is the development of a graph similarity measure to display multiple grid graphs according to their similarity to a query graph. Due to the screen size restriction, only a limited number of graphs could be displayed in a screen. A useful approach thus is to display only those graphs that are similar to a query graph which represents a production plan that have certain production patterns the user is interesting in. We propose a nested string matching method to access the similarity of grid graphs. Let GQ be a query graph and GC be a compared graph. The method first transform or edit points in each row of GC to points in each row of GQ, which is accomplished by a string matching along the horizontal direction. The match results are then employed in the second string matching which transforms rows of GC to rows of GQ along the vertical direction. Experimental case study is provided to show the effectiveness of the proposed method.