基于NSGAII和神经网络的织造车间大规模调度

doi:10.13475/j.fzxb.20220301801

纺织学报 ›› 2023, Vol. 44 ›› Issue (11): 208-215.doi: 10.13475/j.fzxb.20220301801

基于NSGAII和神经网络的织造车间大规模调度

雷钧杰¹^,², 沈春娅¹^,², 胡旭东¹^,²(), 汝欣¹^,², 彭来湖¹^,²

1.浙江理工大学机械与自动控制学院, 浙江杭州 310018
2.浙江理工大学浙江省现代纺织装备技术重点实验室, 浙江杭州 310018

收稿日期:2023-03-04 修回日期:2023-06-24 出版日期:2023-11-15 发布日期:2023-12-25
通讯作者: 胡旭东(1959—),男,教授,博士。主要研究方向为智能纺织装备技术。E-mail:xdhu@zstu.edu.cn。
作者简介:雷钧杰(1995—),男,硕士。主要研究方向为纺织智能调度。
基金资助:
浙江省公益技术研究计划项目(LGG21E050024);浙江省重点研发计划项目(2019C01038);浙江省博士后科研项目特别资助项目(ZJ2020004);浙江理工大学科研启动基金项目(18022224-Y)

Large-scale scheduling of weaving workshop based on NSGAII and neural network

LEI Junjie¹^,², SHEN Chunya¹^,², HU Xudong¹^,²(), RU Xin¹^,², PENG Laihu¹^,²

1. Faculty of Mechanical Engineering & Automation, Zhejiang Sci-Tech University, Hangzhou, Zhejiang 310018, China
2. Key Laboratory of Modern Textile Machinery & Technology of Zhejiang Province, Zhejiang Sci-Tech University, Hangzhou, Zhejiang 310018, China

Received:2023-03-04 Revised:2023-06-24 Published:2023-11-15 Online:2023-12-25

摘要/Abstract

摘要：

为解决遗传算法在织造车间大规模调度中容易陷入局部最优的问题,提出了NSGAII-NN125调度算法。首先,根据织造车间大规模调度的特点,以最小化逾期损失、完工时间和改车次数为优化目标,建立了织造车间调度模型。然后设计了以神经网络模型NN125为主体的调度模块,其可根据织轴和织机特征信息生成调度方案。最后,设计了以NSGAII为主体的优化模块,其根据方案优劣对调度模块中的NN125进行优化。结果表明:NSGAII-NN125的调度质量随着调度规模的不断增大始终非常稳定,而且已优化的调度模块可直接用于相似问题的调度,调度性能较好,由于省去了优化过程,调度速度(约50个织轴/s)也有较大提升,具有较好的实用价值。

关键词: 织造车间, 大规模调度, NSGAII, 神经网络, 多目标调度, 智能调度

Abstract:

Objective With the increase of personnel, machines and materials, the scheduling scale of weaving workshop increases exponentially. The intelligent scheduling algorithm represented by genetic algorithm is easy to fall into the local optimal solution when solving large-scale scheduling problems, and the process is slow, which is difficult to meet the actual demand. This study aims to combine the advantages of genetic algorithm and neural network to solve the problem of large-scale scheduling in weaving workshop.

Method According to the characteristics of large-scale scheduling of weaving workshop, a weaving workshop scheduling model was established to minimize overdue loss, the makespan and the number of variety changes. Then, a weaving workshop scheduling algorithm NSGAII-NN125 based on NSGAII and neural network was proposed to solve the large-scale scheduling problem of weaving workshop, which consists of a scheduling module and a multi-objective optimization module. Finally, the optimization module was adopted to find the best the scheduling module according to the merits and demerits of the generation scheme, leading to the scheduling module with high quality, fast speed and reusable.

Results Comparing the objectives of minimizing overdue loss, the makespan and the number of variety changes, NSGAII-NN125 offered stable performance in a series of weaving workshop scheduling, especially in large-scale scheduling with more than 300 looms and more than 2 000 weaver's beams(Tab. 3). The optimization does not fall into the trend of local optimal solution, and the solution quality is outstanding. Compared with the optimization time, NSGAII-NN125 needed to take longer time to calculate and update the eigenvalues of the neural network. The scheduling speed of NSGAII-NN125 was about 0.67 weaver's beams per second. The NN125 model set was optimized by NSGAII-NN125 according to the scheduling requirements of a weaving workshop which can be used for scheduling similar requirements. Compared with the scheduling objectives, it can be seen that the scheduling quality of the optimized NN125 model set is only slightly weaker than that of NSGAII-NN125, and the time consumption is greatly reduced because the long optimization process is eliminated. The scheduling speed is increased to 50 weaver's beams per second, which has good practical value(Tab. 4).

Conclusion The NSGAII-NN125 scheduling algorithm was divided into scheduling module and optimization module in structure. The scheduling and optimization were decoupled, so that the search space of genetic algorithm was limited to a fixed number of parameters in the neural network model, which solves the problem that the GA is easy to fall into the local optimal solution or even scheduling failure due to the large solution space in the large-scale scheduling. More importantly, NSGAII-NN125 outputs the optimal NN125 model set after solving a certain problem. The network model set can be reused to avoid repeated optimization of similar problems and improve the actual scheduling speed, which has good practical value.

Key words: weaving workshop, large-scale scheduling, NSGAII, neural network, multi-objective optimization, intelligent scheduling

中图分类号:

TS111.8

雷钧杰, 沈春娅, 胡旭东, 汝欣, 彭来湖. 基于NSGAII和神经网络的织造车间大规模调度[J]. 纺织学报, 2023, 44(11): 208-215.

LEI Junjie, SHEN Chunya, HU Xudong, RU Xin, PENG Laihu. Large-scale scheduling of weaving workshop based on NSGAII and neural network[J]. Journal of Textile Research, 2023, 44(11): 208-215.

图/表 7

图1

表1

NN125详细信息"

层编号	神经元数量	单个神经原运算表达式	参数个数
1	5	v=v, t=t,g=g,d=d,p=p	0
2	5	y²^j=ReLU(v $w 1 2 j$ +t $w 2 2 j$ +g $w 3 2 j$ +d $w 4 2 j$ +p $w 5 2 j$ +b³^j)	30
3	5	y³^j=ReLU(y²¹ $w 1 3 j$ +y²² $w 2 3 j$ +y²³ $w 3 3 j$ +y²⁴ $w 4 3 j$ +y²⁵ $w 5 3 j$ +b³^j)	30
4	5	y⁴^j=ReLU(y³¹ $w 1 4 j$ +y³² $w 2 4 j$ +y³³ $w 3 4 j$ +y³⁴ $w 4 4 j$ +y³⁵ $w 5 4 j$ +b⁴^j)	30
5	5	y⁵^j=ReLU(y⁴¹ $w 1 5 j$ +y⁴² $w 2 5 j$ +y⁴³ $w 3 5 j$ +y⁴⁴ $w 4 5 j$ +y⁴⁵ $w 5 5 j$ +b⁵^j)	30
6	1	y⁶=y⁵¹ $w 15$ +y⁵² $w 25$ +y⁵³ $w 35$ +y⁵⁴ $w 45$ +y⁵⁵ $w 55$	5

表1

图2

图3

表2

表3

表4

参考文献 15

[1]	程禹, 赵宏伟, 龙曼丽, 等. 最早截止期优先调度算法的改进[J]. 吉林大学学报(工学版), 2013, 43(5): 1338-1342.
	CHENG Yu, ZHAO Hongwei, LONG Manli, et al. Improvement for earliest deadline first algorithm[J]. Journal of Jilin University (Engineering Edition), 2013, 43(5): 1338-1342.
[2]	KRITHI Ramamritham, JOHN A Stankovic, PERNG-FEI Shiah. Efficient scheduling algorithms for real-time multiprocessor systems[J]. IEEE Trans Parallel Distrib Syst, 1990, 1(2): 184-194. doi: 10.1109/71.80146
[3]	宾雪莲, 杨玉海, 金士尧. 一种基于分组与适当选取策略的实时多处理器系统的动态调度算法[J]. 计算机学报, 2006(1): 81-91.
	BIN Xuelian, YANG Yuhai, JIN Shiyao. A new dynamic scheduling algorithm for real-time multiprocessor systems based on grouping and properly choosing policies[J]. Chinese Journal of Computers, 2006(1): 81-91.
[4]	HE Wentao, MENG Shuo, WANG Jing'an, et al. Weaving scheduling based on an improved ant colony algorithm[J]. Textile Research Journal, 2021, 91(5/6):1-7.
[5]	黄学文, 陈绍芬, 周阗玉, 等. 求解柔性作业车间调度的遗传算法综述[J]. 计算机集成制造系统, 2022, 28(2): 536-551.
	HUANG Xuewen, CHEN Shaofen, ZHOU Tianyu, et al. Survey on genetic algorithms for solving flexible job-shop scheduling problem[J]. Computer Integrated Manufacturing Systems, 2022, 28(2): 536-551.
[6]	顾文斌, 李育鑫, 钱煜晖, 等. 基于激素调节机制IPSO算法的相同并行机混合流水车间调度问题[J]. 计算机集成制造系统, 2021, 27(10): 2858-2871. doi: 10.13196/j.cims.2021.10.010
	GU Wenbin, LI Yuxin, QIAN Yuhui, et al. Improved particle swarm optimization algorithm with hormone modulation mechanism for solving hybrid flow-shop scheduling problem with identical parallel machine[J]. Computer Integrated Manufacturing Systems, 2021, 27(10): 2858-2871.
[7]	陈自豪, 陈松航, 陈豪. 多种群NSGA-II改进算法的大规模柔性作业车间调度问题[J]. 传感器与微系统, 2021, 40(2): 51-54.
	CHEN Zihao, CHEN Songhang, CHEN Hao. Large-scale FJSP based on improved multi-group NSGA-II algorithm[J]. Transducer and Microsystem Technologies, 2021, 40(2): 51-54.
[8]	宋存利. 求解多目标混合流水车间调度的改进NSGAII算法[J/OL]. 计算机集成制造系统: 1-19.[2022-04-19]. http://kns.cnki.net/kcms/detail/11.5946.tp.20201026.1044.022.html.
	SONG Cunli. An improved NSGAII algorithm for hybrid flow shop scheduling problem with multi-objec-tive[J/OL]. Computer Integrated Manufacturing Systems: 1-19.[2022-04-19]. http://kns.cnki.net/kcms/detail/11.5946.tp.20201026.1044.022.html.
[9]	陈革, 杨建成. 纺织机械概论[M]. 北京: 中国纺织出版社, 2011: 80-82.
	CHEN Ge, YANG Jiancheng. Introduction to texile machinery[M]. Beijing: China Textile & Apparel Press, 2011: 80-82.
[10]	沈春娅, 雷钧杰, 汝欣, 等. 基于改进型NSGAII的织造车间多目标大规模动态调度[J]. 纺织学报, 2022, 43(4):74-83.
	SHEN Chenya, LEI Junjie, RU Xin, et al. Multi-objective large-scale dynamic scheduling for weaving workshop based on improved NSGAII[J]. Journal of Textile Research, 2022, 43(4):74-83. doi: 10.1177/004051757304300203
[11]	雷钧杰. 织造车间智能调度算法研究[D]. 杭州: 浙江理工大学, 2022:2-50.
	LEI Junjie. Research on intelligent scheduling algorithm for weaving workshop[D]. Hangzhou: Zhejiang Sci-Tech University, 2022:2-50.
[12]	周明, 孙树栋. 遗传算法原理及应用[M]. 北京: 国防工业出版, 1999: 15-16.
	ZHOU Ming, SUN Shudong. Genetic algorithms: theory and application[M]. Beijing: National Defense Industry Press, 1999: 15-16.
[13]	KALYANMOY Deb, SAMIR Agrawal, AMRIT Pratap, et al. A fast and elitist multiobjective genetic algorithm: NSGA-II[J]. IEEE Trans Evolutionary Computation, 2002, 6(2): 182-197. doi: 10.1109/4235.996017
[14]	SAMEER Agarwal, KALYANMOY Deb. Simulated binary crossover for continuous search space[J]. Complex Systems, 2000, 9(3): 115-148.
[15]	KALYANMOY Deb, MAYANK Goyal. A combined genetic adaptive search (GeneAS) for engineering design[J]. Computer Science and Informatics, 1999, 26(4): 30-45.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

参数	参数值
总织机数	M
各织机车速/(m·h^-1)	U[15,21]
结经时间/h	1
同品种换轴时间/h	3
改车换轴时间/h	4
总织轴数	N
织轴绕长/m	U[2 000,3 000]
织轴所属订单号	U[1,ceil(N/7)]
各订单的品种号	U[1,100]
各订单的权重	U[1.0,1.9]
各订单到达时间/h	0
各订单截止时间/h	U[ceil(7/M)×400, N/M ×200]

组别	规模		调度目标值									算法计算耗时/s
	规模		近视算法			NSGAII_3			NSGAII-NN125			近视算法	NSGAII_3	NN125- NSGAII
	M	N	f₁	f₂	f₃	f₁	f₂	f₃	f₁	f₂	f₃	近视算法	NSGAII_3	NN125- NSGAII
	6	50	0	1 306.8	98	0	1 172.6	19	0	1 159.1	26	0.01	4.09	78.98
a	6	100	0	2 537.9	199	0	2 367.8	44	0	2 332.1	32	0.01	7.58	137.82
	6	150	0	3 695.4	294	0	3 495.5	72	0	3 441.8	27	0.01	11.39	251.78
	50	400	0	1 462.6	793	0	1 367.3	298	0	1 166.4	162	0.03	28.40	620.65
	50	700	0	2 292.1	1 348	0	2 316.3	631	0	2 009.7	243	0.05	48.88	1 073.10
b	50	1 000	0	3 274.3	1 930	0	3 254.8	814	0	2 909.2	310	0.07	68.04	1 523.75
	300	2 000	0	1 292.7	3 683	8.3	1 430.9	1 827	0	991.9	1041	0.18	139.22	3 006.80
c	300	4 000	0	2 255.8	7 399	2 081.2	2 675.4	3 859	0	1 918.4	626	0.33	306.42	5 773.55
	300	6 000	0	3 234.2	11 055	2 071.2	3 847.5	5 700	0	2 858.3	1 606	0.48	574.04	8 667.18

组别	M	N	f₂/h	f₃	耗时/s
a	6	50	1 177.71	18	1.07
	6	100	2 361.60	28	1.94
	6	150	3 479.76	57	2.72
b	50	400	1 201.88	160	7.46
	50	700	2 028.71	233	13.31
	50	1 000	2 933.11	364	19.97
c	300	2 000	1 020.94	632	48.76
	300	4 000	1 936.35	996	97.80
	300	6 000	2 858.81	1 606	145.72

基于NSGAII和神经网络的织造车间大规模调度

Large-scale scheduling of weaving workshop based on NSGAII and neural network

RichHTML

PDF (PC)

摘要/Abstract

引用本文

使用本文

图/表 7

参考文献 15

相关文章 15

Metrics

本文评价

推荐阅读 0

[1]	任国栋, 屠佳佳, 李杨, 邱子安, 史伟民. 基于轻量化网络和知识蒸馏的纱线状态检测[J]. 纺织学报, 2023, 44(09): 205-212.
[2]	姚琳涵, 张颖, 姚岚, 郑晓萍, 魏文达, 刘成霞. 基于多尺度纹理合成的刺绣风格迁移模型[J]. 纺织学报, 2023, 44(09): 84-90.
[3]	付晗, 胡峰, 龚杰, 余联庆. 面向织物疵点检测的缺陷重构方法[J]. 纺织学报, 2023, 44(07): 103-109.
[4]	李学良, 杜玉红, 任维佳, 左恒力. 基于近红外光谱和残差神经网络的异性纤维分类识别[J]. 纺织学报, 2023, 44(05): 84-92.
[5]	王斌, 李敏, 雷承霖, 何儒汉. 基于深度学习的织物疵点检测研究进展[J]. 纺织学报, 2023, 44(01): 219-227.
[6]	张福沐, 刘端武, 胡跃明. 染色机助剂智能配送系统的构建及实践[J]. 纺织学报, 2022, 43(11): 179-187.
[7]	刘锋, 徐杰, 柯文博. 基于深度强化学习的服装缝制过程实时动态调度[J]. 纺织学报, 2022, 43(09): 41-48.
[8]	金守峰, 侯一泽, 焦航, 张鹏, 李宇涛. 基于改进AlexNet模型的抓毛织物质量检测方法[J]. 纺织学报, 2022, 43(06): 133-139.
[9]	沈春娅, 雷钧杰, 汝欣, 彭来湖, 胡旭东. 基于改进型NSGAII的织造车间多目标大规模动态调度[J]. 纺织学报, 2022, 43(04): 74-83.
[10]	夏明, 李沛儒, 丁蓉. 基于要素匹配的连衣裙成本快速核算方法[J]. 纺织学报, 2022, 43(03): 146-152.
[11]	江学为, 田润雨, 卢方骁, 张艺. 基于模拟评分的服装推荐改进算法[J]. 纺织学报, 2021, 42(12): 138-144.
[12]	江慧, 马彪. 基于服装风格的款式相似度算法[J]. 纺织学报, 2021, 42(11): 129-136.
[13]	庹武, 王哓玉, 高雅昆, 于媛媛, 郝潇潇, 刘永亮, 郭鑫. 基于改进边缘检测算法的服装款式识别[J]. 纺织学报, 2021, 42(10): 157-162.
[14]	许雪梅. 基于模拟退火算法改进遗传算法的织物智能配色[J]. 纺织学报, 2021, 42(07): 123-128.
[15]	周亚勤, 王攀, 张朋, 张洁. 纬编织造车间生产调度方法研究[J]. 纺织学报, 2021, 42(04): 170-176.