مدل هوشمند تصمیم گیری چند هدفه با تکنولوژی RFID برای برنامه ریزی تولید
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|5887||2013||45 صفحه PDF||سفارش دهید||9800 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : International Journal of Production Economics, Available online 23 May 2013
A multi-objective production planning problem in the labor-intensive manufacturing industry is investigated. An intelligent and real-time multi-objective decision-making model is developed to provide timely and effective solutions for this problem by integrating RFID technology with intelligent optimization techniques, in which RFID technology is used to collect real-time production data, a novel (μ/ρ+λ)(μ/ρ+λ)-evolution strategy process with self-adaptive population size and novel recombination operation is proposed and integrated with effective non-dominated sorting and pruning techniques to generate Pareto optimal solutions for real-world production. Experiments based on industrial data were conducted to evaluate the effectiveness of the proposed model. Experimental results show that the proposed model can effectively solve the investigated problem by providing production planning solutions superior to industrial solutions.
Effective production planning is crucial to delivering products on time to meet customer needs, which can greatly affect the overall performance of a manufacturing enterprise and thus the entire supply chain management. This paper investigates a real-world production planning problem, multi-objective order allocation (MOA), with the consideration of multiple plants and multiple production departments. 1.1. Difficulty in production decision-making practice In today’s labor-intensive manufacturing industries, production data are usually collected by using barcode technology or even manual input in an offline manner. The data collected are never real-time and their accuracy is questionable, which hinders the research and application of effective methodologies in production planning and control since the accuracy and real-timeness of production data are the premises of implementing effective production decision-making. In recent years, radio-frequency identification (RFID) has attracted increasing industrial applications as an alternative to the barcode technology (Wyld, 2006), which utilizes electromagnetic wave communication to exchange data between a terminal and an electronic tag attached to an object for identification and tracking. As the application of RFID has become economically feasible, some commercial RFID-based data capture systems, such as ZymFactory system (Zymmetry Group, 2007) and GPRO system (GPROTechnologies, 2010), have been developed to obtain real-time and accurate production data and their effectiveness has been proved by various industrial applications and practices in labor-intensive manufacturing (Guo, 2008). However, these systems only focus on data collection and simple data reporting, and cannot provide decision-making solutions to assist production managers in performing production management. In today’s labor-intensive manufacturing industries (such as apparel and footwear), manufacturing is characterized by short production lead-time, tight delivery due dates, small quantities with frequent product change, as well as the multi-plant and multi-production department nature. These phenomena and characteristics increase the complexity of making effective production decisions. Decision-making on production planning in these industries relies heavily on the production planners’ experience and subjective assessments, which may not be consistent under similar conditions and is thus non-optimal. 1.2. Previous studies in production planning In the production planning area, a great number of papers have been published and there exist some comprehensive review papers (Dolgui and Prodhon, 2007, Wang et al., 2009 and Wazed et al., 2010). Research issues in this area mainly include master production schedule (Sahin et al., 2008), material requirements planning (Dolgui and Prodhon 2007), manufacturing resource planning (Wazed et al., 2010) and aggregate planning (Al-E-Hashem et al., 2011). To assist manufacturers in assigning production processes of each order to appropriate plants, some researchers investigated order allocation and release problems in production planning stage, which are an important decision-making problems in labor-intensive industries because its performance greatly affects that of downstream production control and the entire supply chain. Ashby and Uzsoy (1995) proposed a set of heuristic rules to integrate order release, group scheduling and order sequencing into a single-stage production system. Chen et al. (2005) presented a decision support system to determine how to assign a particular order to the most appropriate manufacturing company from the global supply chain’s perspective. Work by Axsater (2005) investigated the order release problem in a multi-stage assembly network by determining the starting time of different production operations. Chen and Pundoor (2006) investigated order allocation and scheduling at the supply chain level by assigning orders to different plants and exploring a schedule to perform the orders assigned in each plant. However, their study has not considered the effects of different production departments in each plant on production decision-making performance. The order allocation problem in the production planning stage, with the consideration of multi-plant and multi-production department features, has not been addressed so far. Unfortunately, these features are typical in labor-intensive manufacturing, which significantly increases the complexity of production planning problems. The MOA problem investigated in this research is a computationally complex combinatorial optimization problem because it handles the assignments of production processes of multiple orders to production departments of multiple plants, which lead to a huge solution space. 1.3. Techniques for optimization problems in production decision-making To obtain effective solutions to optimization problems in production decision-making, a wide variety of techniques have been developed (Guo, 2008 and Guo et al., 2011), mainly including simulation-based techniques (Chan et al., 2002), priority-rule-based techniques (Weng and Ren, 2006), classical optimization techniques (Tanaka and Araki, 2008 and Tozkapan et al., 2003), and meta-heuristic techniques. The first three types of techniques cannot provide effective solutions to complex optimization problems in real-world production usually due to the high complexity of such problems. The meta-heuristic techniques have been proved to be very powerful in finding optimal or near-optimal solutions due to their heuristic nature (Blum and Roli, 2003 and Luna et al., 2010). The most commonly used meta-heuristic techniques are evolutionary algorithms, especially genetic algorithm (Holland 1975) and evolution strategy (ES) (Schwefel, 1995). Comparing with genetic algorithms, the applications of ES in production optimization problems have attracted relatively little attention. It is usual that multiple optimization objectives, some of which are in conflict, need to be achieved simultaneously in many real-world optimization problems. Some researchers use the weighted sum method to turn the multi-objective problems to single-objective ones (Guo et al., 2008 and Ishibuchi and Murata, 1998). However, it is impossible to have a single solution which can simultaneously optimize all objectives when multiple objectives are conflicting. To handle this problem, some researchers developed multi-objective optimization algorithms by introducing the Pareto optimality concept into the meta-heuristic techniques so as to provide more feasible solutions (i.e., Pareto optimal solutions). The most well-known ones include NSGA-II (Deb et al., 2002), PAES (Knowles and Corne, 2000), SPEA2 (Zitzler et al., 2001), IBEA (Zitzler and Kunzli, 2004), MSOPS (Hughes, 2005). However, these algorithms have not been reported to handle combinatorial optimization problems in production planning. The existing multi-objective optimization algorithms cannot be directly used to handle the MOA problem because different solution representations and evolutionary operators are probably required to handle various problem-dependent features. Moreover, Eiben et al. (2004) reported that self-adaptive population size adjustment has the potential to improve the evolutionary speed of single-objective optimization processes. However, its effectiveness on multi-objective optimization processes has not been investigated. In this paper, an intelligent and real-time multi-objective decision-making (IRMD) model, which integrates an RFID-based production data capture (RPDC) submodel with a heuristic data extraction and analysis (HDEA) submodel and a novel ES-based Pareto optimization (ESPO) submodel, is developed to provide timely and effective solutions for the MOA problem investigated. To construct the ESPO submodel, the ES process with self-adaptive population size and novel recombination operation is proposed and integrated with effective non-dominated sorting and pruning techniques so as to generate Pareto optimal production planning solutions. The rest of this paper is organized as follows. Section 2 formulates the investigated production planning problem. In Section 3, the IRMD model is presented to handle this problem. In Sections 4 and 5, experimental comparisons and analyses are conducted to validate the effectiveness of the proposed model. Finally, this paper is summarized and future research directions are suggested in Section 6.
نتیجه گیری انگلیسی
In this paper, we investigate a production planning problem in labor-intensive manufacturing industries, the MOA problem with the consideration of multiple plants and multiple production departments, which aims at assigning production processes of each order to appropriate plants. A novel IRMD model was developed to deal with the investigated problem, which consists of a RPDC submodel, a HDEA submodel and an ESPO submodel. The RPDC submodel captures necessary production data for further data extraction and optimizing processes from production workstations by integrating commercial RPDC systems. The HDEA submodel extracts necessary data from production data collected by the RPDC submodel as the inputs of the ESPO submodel. The ESPO submodel constructs a novel optimum-seeking process to obtain effective order allocation solutions for real-world production by integrating a multi-parent recombination operator, a faster non-dominated sorting technique, a self-adaptive population size adjustment method and a solution pruning technique into the (μ/ρ+λ)--ES process. Real-time production data from industrial practice were obtained based on the RPDC submodel, and utilized to validate the proposed IRMD model. Experimental results demonstrated that the proposed model can solve the MOA problem effectively by providing Pareto optimal solutions superior to the industrial solutions. The self-adaptive population size adjustment during the evolutionary process is helpful to improve the optimum-seeking performance of multi-objective optimization processes. This research does not compare the IRMD model developed with others in the area of production planning research since similar research has not yet been published. Further research will focus on the improvement of the methodology to solve MOA problems with more production objectives and practical constraints, and on the effects of various uncertainties on production planning, including uncertain production orders and possible material shortage.