An Inventory Replenishment Model for Deteriorating Items with Time-varying Demand and Shortages using Genetic Algorithm
Abstract
In this paper, an inventory replenishment model for deteriorating items is developed. Demand for the item varies with time over a finite planning horizon, during which shortages are allowed and are completely back-ordered. The objective is to determine a replenishment policy that minimizes the total inventory cost. A search procedure based on Genetic Algorithm (GA) is presented and illustrated with some numerical results.
Keywords: Inventory; Replenishment; Time-varying demand; Deterioration; Shortages; Genetic Algorithms
1. Introduction
When developing inventory models, many simplifying assumptions are made to find the optimal replenishment policies. The Economic Order Quantity (EOQ) model assumes a constant demand rate over an infinite planning horizon and minimizes the total inventory cost per unit of time. However, practical demand rates rarely behave this way. Most items experience stable demand only during the saturation phase of their life cycle, while demand is increasing or decreasing during their growth or decline phases respectively. In addition, most models assume that items have infinite shelf-life while in storage. This assumption, while applicable to items with low deterioration rates, seems unrealistic for volatile and radioactive materials, blood banks, food stuff, electronics, etc, that continually lose their utility while in stock. Shortages may be economically desirable in many situations such as when storage costs are high compared to back-order costs or when storage space is limited, is another important aspect that needs to be given special attention.
Hariga (1994) developed an optimal procedure to optimize an inventory model with time-varying demand and shortages. A similar procedure was adopted by Hariga and Benkherouf(1994) to optimize an inventory model with time-varying demand and deterioration, but without shortages. Later, Benkherouf(1995)extended the latter model to include shortages as well. Although this model considered only decreasing demands, it was mentioned that a slight modification would allow the optimal schedule for increasing demands to be found. Other contributors to this model include Wee(1995), Hariga and Al-Alyan (1997), Giri et al.(2000), and Chu and Chen (2002).
Most literatures on this model, as far as the authors are aware of, considered only analytical or heuristical approaches to directly find or approximate its optimal solution. In this paper, we shall consider a stochastic approach by using Genetic Algorithm (GA). Our model operates over a finite planning horizon to satisfy a time-varying demand and assumes that deterioration occurs at a constant rate and shortages are allowed and are completely back-ordered. The objective is to determine a replenishment policy that minimizes the total inventory cost.
This paper is organised as follows: In the next section, we present the mathematical formulation of our model. In Section 3, we describe a GA-based search procedure to approximate the optimal solution. In Section 4, we illustrate this procedure with some numerical examples. We also compare our results with some optimal and heuristical results. Finally, Section 5 summarizes our findings followed by appendices describing an optimal and several heuristical procedures.
2. Mathematical Formulation
The following assumptions are made in developing our mathematical model:
- A single type of item is held in stock over a finite planing horizon H units of time long.
- Replenishment occurs at an infinite rate with zero lead time and is charged with a cost of per replenishment.
- Inventory holding cost is charged only to good units with a cost of per unit per unit of time.
- Shortages are allowed and are completely back-ordered. Units short are charged with a cost of per unit per unit of time until they are cleared by back-orders.
- The item deteriorates at a fixed rate theta;. The deterioration of the units is considered only after their receipt into storage and the deteriorated units are not replaced or repaired during the planing horizon, but are charged with a cost of per unit.
- The demand rate is a continuous, deterministic and time-varying function and gt; 0 for 0 le; t le; H.
Suppose that n replenishment (including the final backorder) are made during the planning horizon. Also, suppose that the total time elapsed up to and including the jth replenishment cycle (j=1,2,hellip;,n-1) is given by . We define =0 and =. Moreover, suppose that the total time elapsed up to the start of shortage in the jth cycle is given by .
To start with, consider the jth cycle. Initially, the stock is zero. Replenishment occurs at and the inventory level reaches its maximum immediately. The positive inventory level is represented by . This amount of inventory is depleted over time by demand and deterioration until it reaches zero at() . Now, a shortage occurs and the negative inventory level is represented by. The shortage peaks at and is immediately cleared by a back-order. The cycle then repeats itself. We note that the back-order made during the cycle (,) and the replenishment made during the cycle (,) incur a single replenishment cost since they both happen at the same time.
Hence, the total inventory cost, which is defined as the sum of replenishment, holding, deterioration and shortage costs, is given by
(1)
The variations of and with respect to time in the (,) cycle is governed by the following linear differential equations:
, (2)
, (3)
With the boundary conditions and
全文共15656字,剩余内容已隐藏,支付完成后下载完整资料
一种时变需求和基于遗传算法短缺的变质品库存补货模型
摘要
本文提出了一种变质品的补货模型。物品需求随时间在一个有限的计划期内不断变化,在此期间,允许缺货并完全拖后。我们的目的是确定补货策略使得总库存成本最小。提出了一种基于遗传算法的搜索过程,并给出了一些数值结果。
关键词:库存;补货;时变需求;变质;短缺;遗传算法
1. 简介
在建立库存模型时,为了找到最优的补货策略作出许多简单化的假设。经济订货批量公式(EOQ)模型假定一个超过无限计划期的常数需求率,最大限度地减少每单位时间的总库存成本。然而,实际需求率很少是这样的方式。大多数物品只有在它们的生命周期饱和阶段经历了稳定的需求,而需求在它们生产期或衰退期阶段是增加或减少的。此外,大多数的模型假设物品在仓库中有无限的保存期限。这种假设,同时适用于在库存中不断失去效用的低变质率的变质物品,如看起来不切实际的挥发性和放射性物质、血液、食品、电子等。另一个重要的方面,在很多的情况下短缺在经济上可能是可取的,例如:当存储成本与拖后费用相比很高或者当存储空间有限,需要给予特别的关注。
Hariga(1994)发明了一种最优程序来优化时变需求和允许短缺的库存模型。一种类似的程序被Hariga和Benkherouf (1994) 采用来优化时变需求和不允许短缺的变质物品的库存模型。后来,Benkherouf(1995)推广了后一种模型,使其允许缺货。虽然该模型只考虑到需求逐渐减少,它却提到一个轻微的修改就可以使最佳安排中日益增加的需求被发现。该模型的其他贡献者,包括Wee (1995)、Hariga和Al-Alyan (1997)、 Giri(2000)、Chu与Chen(2002)等人。
大多数对该模型的研究文献中,作者们意识到,只考虑途径分析或搜索方法就可以直接或间接找到其最优解。在本文中,我们将考虑一种利用遗传算法(GA) 的随机方法。我们的模型运行在有限计划期内来满足一种时变需求和假设变质现象以恒定的速率发生和允许缺货并完全拖后的情况。这样做的目的是确定使总库存成本最小的补货策略。
本文安排如下:在下一节中,我们介绍了模型的数学公式。在第三节中,我们描述了一种基于遗传算法的搜索过程求解近似最优解。在第四节,我们举一些数值例子说明该程序。我们把所得结果与一些搜索到的最优结果进行比较。最后,第五节总结我们的研究,紧随其后的是附录,描述一个最优程序和几个搜索程序。
2. 数学公式
下面是我们数学模型所作的假设:
- 对一个单一类型的物品,在一个有限的计划期内存贮在仓库中,计划时域长度为H。
- 瞬时补货,提前期为零,每次的订货固定费用为;
- 库存持有成本为物品的存贮费,每单位物品单位时间的费用为;
- 允许缺货并完全拖后。每单位物品单位时间的缺货费用为,直到他们通过延期交货补足;
- 物品的变质率为常数,变质品在进入仓库之后开始变质,并且变质品在计划期内不能进行替代或修复,每单位的物品变质损失费为;
- 需求率是连续的、确定的和随时间变化的函数,且gt; 0
(0 le; t le; H)。
假设补货数量为(包括最后的缺货),在计划期内完成。此外,假设第个补货周期内开始补货的时刻点为 (j=1,2,hellip;,n-1)。我们定义=0 和=。并且,假设在第个补货周期内开始缺货的时刻点为。
首先,考虑第j个周期的情况。最初,库存量为零。补货发生在时刻,库存水平立刻达到最大。具体的库存水平用表示。随着需求和物品变质,库存数量逐渐减少,直到()时刻降为零为止。此时,缺货发生,负的库存水平用表示。短缺量在时刻达到顶峰,直接通过延期交货补足。之后,订货周期反复循环。我们注意到在(,)周期内的缺货和在(,)周期内的补货仅引起一个补货费用,因为他们发生在同一时间。
因此,被定义为补货费用、存贮费用、变质损失费用和短缺费用之和的总库存成本为:
(1)
在(,)周期内和随时间的变化满足下面的线性微分方程:
, (2)
, (3)
有边界条件和。式子(2)和(3)的解为:
, (4)
, (5)
随后,在(,)周期内的库存量为:
(6)
在(,)周期内对于短缺数量,我们结合在时间间隔内的表达式得到:
(7)
n-补货策略的总库存成本为:
(8)
其中。
我们的问题是寻找最佳补货策略使最小,n和最佳补货点服从以下限制:
,n是整数,是实数。
对于固定的n,最佳的和可以从下面的一组方程得到:
, (9)
, (10)
注意,默认情况下,=0和=。
用遗传算法程序确定最佳的,根据得到:
(11)
3.遗传算法(GA)搜索程序
遗传算法(GA)是一种模仿自然生物进化的随机搜索方法。GA探究了保持个体的种群问题的领域,每一问题域代表一个可能的解决这个问题的方法,直到满足某些停止标准为止。对于这种类型的库存模型,在大多数的文献中找到最优或近似最优解的解决方法是确定一个固定的n,而最优的n是通过评价一系列的W(n)得到的。然而,我们的遗传算法程序将把n看作一个变量并且把除了补货点以外的n演变。
给在一个种群中的每个个体编码时,n的值是整数并且它的相关实时补货点()满足下列条件:
,。
值得注意的是补货点的数量是依赖n的,因此不是固定的。所以,这导致了个体的种群是可变长度的。为了创建初始种群,我们随机生成一系列的n,其范围是任意设定的,然后为每个n设定适当的补货数量。我们设第一个个体有n = 2,且无补货点代表了延期交货的单一补货策略。因此,个体矩阵采取以下形式
(12)
其中NaN是一个非数字常量,不参与到程序中去。
我们利用(8)式评估一个种群的目标价值。观察到,在前一节描述的数学模型中,总库存成本是一个关于() 和 的函数,本身是关于()的函数,其中最优值能从每一对通过解(11)式得到的()得到。我们在搜索过程中利用随机抽样方法(Baker,1987),利用非线性排序法给每个个体分配适合度。
我们限制交叉只能在长度相同的个体之间进行,类似的多个种群的遗传算法凭借仅限于在相同的筛选里的那些个体交换遗传信息。我们选择了离散重组算子(Chipperfield等,1993年)完成了交叉。
我们做的突变过程分为两个阶段。在第一阶段,我们用一个会突变成任何大于2但有一个固定的上限的整数的突变概率表现关于整数n的突变概率。从补货点的数量是依赖于n的,我们设计一个程序,当n增大,然后随机产生的额外补货点会相应的增加。然后,我们按升序整理成新的补货点集。如果n减少,那么我们从集合删除最后一个补货点。这就消除了重新整理设置的需要。我们有限的实验结果表明,该方法并不能对该算法的收敛性产生巨大影响。在第二阶段,我们表示出利用突变遗传算法的补货点 (Muuml;hlenbei和Schierkamp-Voosen,1993)。我们变异每一个有1/( n-2)概率的订货点,其中(n-2)相当于关于每个个体的补货点的数量。
4. 数值例子
我们的程序是使用Matlab高级语言编写的,是一个运行在奔腾4处理器兼容IBM,功率1.5千兆赫的机器上的一个Matlab 6.1环境。在我们所有的实验中,下列值是用于GA的参数(Chipperfield等,1993年)的:种群数量= 30,基因差距= 0.9,重组率= 0.9,n突变率= 0.5,插入率= 0.9,强制选择= 2,同群个体的最大数量= 1000。
我们利用四个实例对我们的程序加以说明:两个需求函数为的线性需求,两个需求函数为 的指数需求。这些实例的参数值展示在表1中。对于每个实例,我们运行10次我们的程序,结果展示在表2中。表3给出了最优程序的最佳W值(精确到小数点后5位)和两种搜索程序得到的值;这些在附录中描述。
表1: 实例的参数值
需求情况 |
a |
b |
H |
c1 |
c2 |
c3 |
c4 |
theta; |
线性 1 |
6 |
1 |
11 |
9 |
0.1 |
0.3 |
0.2 |
0.04 |
线性 2 |
100 |
-10 |
4.78 |
1.5 |
0.2 |
0.8 |
0.4 |
0.04 |
指数 1 |
500 |
-0.98 |
11 |
9 |
0.1 |
0.3 |
0.2 |
0.04 |
指数 2 |
0.1767 |
1 |
7 |
1.5 |
0.2 |
0.8 |
0.4 |
0.04 |
表 2: 使用遗传算法程序得到的实例结果
需求情况 |
最优 W |
最劣 W |
平均 W |
线性 1 |
55.04 |
55.04 |
55.04 |
线性 2 |
31.13 |
31.34 |
31.20 |
指数 1 |
59.27 |
59.27 |
59.27 |
指数 2 |
21.79 |
22.49 |
21.93 |
表 3: 使用最优程序和5种搜索方法得到的实例结果
需求情况 |
线性 1 |
线性 2 |
指数 1 |
指数 2 |
最优法 |
55.04 |
31.13 |
59.27 |
21.79 |
搜索法 1 |
59.87 |
31.25 |
89.08 |
3 全文共5603字,剩余内容已隐藏,支付完成后下载完整资料 资料编号:[15563],资料为PDF文档或Word文档,PDF文档可免费转换为Word |
以上是毕业论文外文翻译,课题毕业论文、任务书、文献综述、开题报告、程序设计、图纸设计等资料可联系客服协助查找。