﻿ 面向动用计划的车辆装备备件预测研究
«上一篇
 文章快速检索 高级检索

 智能系统学报  2021, Vol. 16 Issue (6): 1064-1072  DOI: 10.11992/tis.202012026 0

### 引用本文

LI Fengyue, QI Xiaogang, BAN Liming, et al. Vehicle maintenance spare-part prediction for equipment use plan[J]. CAAI Transactions on Intelligent Systems, 2021, 16(6): 1064-1072. DOI: 10.11992/tis.202012026.

### 文章历史

1. 西安电子科技大学 数学与统计学院，陕西 西安 710000;
2. 中国人民解放军32272部队，甘肃 兰州 730000

Vehicle maintenance spare-part prediction for equipment use plan
LI Fengyue 1, QI Xiaogang 1, BAN Liming 2, LI Jianhua 2, SUO Wenkai 2
1. School of Mathematics and Statistics, Xidian University, Xi’an 710000, China;
2. 32272 Group of PLA, Lanzhou 730000, China
Abstract: Aiming at the consumption characteristics of vehicle equipment spare parts under the use plan, two forecast optimization problems of spare-part consumption and inventory control are studied. First, the preventive and repairable maintenance of vehicle equipment is considered for the use plan period, and the consumption of equipment maintenance spare parts under the fixed schedule and random natural failure maintenance is predicted. Subsequently, based on the characteristics of the spare parts inventory inspection method, the joint replenishment inventory control model is established based on a regular inspection strategy. According to the structural characteristics of the model, the limits of decision variables are determined, and the multipopulation location upgrade method is used to improve the fruit fly optimization algorithm (FOA). The simulation results show that the improved FOA has good solving efficiency. The proposed optimization method can provide a decision-making basis for the optimization of vehicle maintenance support resources.
Key words: vehicle equipment    spare parts prediction    using plan    consumption    regular inspection    inventory control    multi-population    fruit fly optimization algorithm

1 车辆装备备件消耗量模型 1.1 车辆装备备件消耗过程分析

1.2 车辆装备备件消耗量模型

1)定时定程维修备件消耗量

 ${Z_j} = \sum\limits_{k = 1}^N {\left\lfloor {\frac{{{M_k} - {M_{jk}} + {{\rm{EM}}_k}}}{{{{\rm{IM}}_j}}}} \right\rfloor }$ (1)

 ${D'_i} = \sum\limits_{\forall f} {\sum\limits_{j = 1}^3 {{Z_j} \cdot } {\gamma _{jf}} \cdot {c_{fi}}}$ (2)

2)随机故障维修备件消耗量

 ${H_f} = \sum\limits_{\forall l} {\sum\limits_{\forall k} {\Delta {M_{kl}}{\lambda _l}{\beta _f}} }$ (3)

 ${D''_i} = \sum\limits_{\forall f} {({H_f} \cdot {c_{fi}})}$ (4)

3)动用计划下备件消耗量模型

 $\begin{array}{l} {D_i} = {D'_i} + {D''_i} = \displaystyle\sum\limits_{\forall f} {\displaystyle\sum\limits_{j = 1}^3 {{Z_j} \cdot } {\gamma _{jf}} \cdot {c_{fi}}} + \displaystyle\sum\limits_{\forall f} {({H_f} \cdot {c_{fi}})} \end{array}$ (5)
2 车辆装备备件库存控制模型

$N$ 种装备备件中，第 $i$ 种备件的订购量 ${Q_i}$

 ${Q_i} = {D_i}{k_i}T, \;\;i = 1,2, \cdots, N$ (6)

 ${C_h} = \frac{1}{2}\sum\limits_{i = 1}^N {{Q_i}{h_i}} = {{\left( {\sum\limits_{i = 1}^N {{k_i}T{D_i}} } \right)} / 2}$ (7)

 ${C_r} = \left( {{R / T}} \right) + \sum\limits_{i = 1}^N {\left( {{{{r_i}} / {{k_i}T}}} \right)}$ (8)

 $\begin{array}{l} C(T,K) = {C_h} + {C_r} = \displaystyle\sum\limits_{i = 1}^N {\dfrac{{{k_i}T{D_i}}}{2}} + \dfrac{R}{T} + \displaystyle\sum\limits_{i = 1}^N {\dfrac{{{r_i}}}{{{k_i}T}}} \end{array}$ (9)

3 算法设计

3.1 界限确定

 ${T^*}(K) = \sqrt {{{2\left( {R + \sum\limits_{i = 1}^N {\frac{{{r_i}}}{{{k_i}}}} } \right)} / {\sum\limits_{i = 1}^N {{k_i}{D_i}{h_i}} }}}$ (10)

 $C(K) = \sqrt {2(R + \sum\limits_{i = 1}^N {\frac{{{r_i}}}{{{k_i}}}} )\sum\limits_{i = 1}^N {{D_i}{k_i}{h_i}} } {\text{ }}$ (11)

${k_i}$ 为大于零的整数，所以显然可以设置下限 ${k_{{\rm{LB}}i}}$ =1。另一方面，通过 $\left( {{{\partial C} / {\partial {k_i}}}} \right) = 0$ 和基本不等式(12)，易得出不等式(13)：

 ${k_i}({k_i} - 1) \leqslant k_i^2 \leqslant {k_i}({k_i} + 1)$ (12)
 ${k_i}({k_i} - 1) \leqslant \frac{{2{r_i}}}{{{D_i}{h_i}{T^2}}} \leqslant {k_i}({k_i} + 1)$ (13)

 ${k_{{\rm{UB}}i}}({k_{{\rm{UB}}i}} - 1) \leqslant \frac{{2{r_i}}}{{{D_i}{h_i}T_{\min }^2}} \leqslant {k_{{\rm{UB}}i}}({k_{{\rm{UB}}i}} + 1)$ (14)

3.2 种群更新方式

1)果蝇“发现者”位置更新方式

 $X_y^{t + 1} = X_y^t + {\text{rand}}({\text{F}}{{\text{R}}_1})$ (15)

2)果蝇“跟随者”位置更新方式

 $X_y^{t + 1} = X_d^{t + 1} + {\text{rand}}\left( {{\text{F}}{{\text{R}}_2}} \right)$ (16)

3)果蝇“反捕食行为”方式

 $X_y^{t + 1} = X\_{\text{axis}} + {\text{rand}}\left( {{\text{F}}{{\text{R}}_3}} \right)$ (17)

1)初始化参数。设置种群规模、最大迭代次数、飞行方向，果蝇各类群体比例。随机初始化果蝇个体位置。

2)计算果蝇的气味浓度值。气味浓度判断值 ${S_y} = {X_y}$ ，将 ${S_y}$ 代入味道浓度判定函数 $F(X)$ 中，本文的判定函数 $F(X)$ 为总成本函数 $C(K)$ 。计算每个个体当前位置的味道浓度值 ${\text{Smel}}{{\text{l}}_y}$ ${\text{Smel}}{{\text{l}}_y} = F({S_y})$

3)对气味浓度值进行排序 ${\text{sort}}({X_i})$ 。记录种群的全局历史最佳位置 $X\_{\text{axis}}$ 和最佳气味浓度值 ${\text{Smellbest}}$ ；根据气味浓度值大小和群体比例参数 $M$ 设置果蝇“发现者”“跟随者”和“反捕食行为”群体。

4)多类种群位置更新。根据式(15)~(17)更新果蝇群体位置。

5)重复步骤2)~4)，直到迭代次数最大，停止迭代并输出结果。

4 仿真与分析 4.1 实例仿真

4.2 仿真分析