UBC Theses and Dissertations
Adaptive optimization methods to improve the performance of chop saw systems Ye, Lin
Optimizing chop saw is one of the primary methods for adding value to rough lumber (boards) and it is common to many breakout lines for plants processing solid wood. Because of the variety of boards to be chopped in real time and the diversity of parts (products) required on a cutting list, there are numerous ways of cutting the boards. The purpose of this thesis was to advance the optimization algorithms that can be used in real-time to generate optimal cutting patterns, aiming to minimize the overall cost of the chop saw operation. Adaptive methods were used in the models to acquire the knowledge on the quality of boards and the inventory level for each part in real-time. The first part of the project developed a real time goal-seeking model for chop saw optimization systems that can be implemented as an on-line control system. Two prioritizing strategies were used in the model by applying an adaptive method that can link the part value with the production level for each part during the production process. A simulation study was conducted to compare the performance of the goal-seeking model with a published model. The result shows that by selecting a proper prioritization strategy, the goal-seeking model can achieve a lower level of overproduction, higher part yield and lower overall cost. The study also found that the goal-seeking model balances the production rates for each part on the cutting list over the production run. The second part of the project developed a combined linear programming and dynamic programming model that used adaptive optimization methods. This model considers the part size and quantity in the cutting list and the grade of boards being processed and the production level for each part chopped in real-time. The simulation study shows that, different from goal-seeking model that requires selecting a proper prioritization strategy for a given cutting list, the combined model has a steadier control to overproduction and good performance against different cutting lists. However, the result is sensitive to the overproduction penalty assigned to the model.
Item Citations and Data