| Automotive industry is one of the manufacturing industries featuring large production and sales volume,complex products,labor-intensive,long supply chain and strong industry-boosting ability.It occupies an important position in the national economic development.Under the guidance of “Made in China 2025” and other strategies,automotive manufacturers are currently carrying out intelligent manufacturing upgrades.Parts logistics is an important link of automobile production,which covers demand information transmission,picking up and collecting goods,transportation and handling,and storage and distribution,etc.It is characterized by multiple link spans,complex categories and large optimization space,etc.Building lean intelligent parts logistics is an important way for automobile manufacturers to realize intelligent manufacturing.This dissertation focuses on the lean intelligent parts logistics of automotive manufacturers.Based on the real production scenario of automobile manufacturers in the context of the impact of COVID-19,this dissertation combines the demand side and supply side,proposes a lean intelligent parts logistics model for the whole cycle of automobile supply chain and conducts an optimization analysis for the whole life cycle of physical flow and information flow of logistics.The research basically realizes the closed-loop study from obtaining demand from dealers,generating parts information from manufacturers,purchasing parts from suppliers,assembling parts from manufacturers,and completing assembly.The research covers the main links of parts logistics,including complex market mining and prediction,optimal balance of assembly line,lean pickup of procurement logistics and unmanned intelligent distribution.The main content is as following:Firstly,accurate grasp of market demand is the starting point for carrying out lean intelligent parts logistics,especially since the outbreak of COVID-19 pandemic disrupted the order of the automobile consumption market.In order to tap the real will of consumers and grasp the real-time market demand under the complex market environment,this dissertation collects daily production and sales data of two automotive manufacturers during the pandemic,domestic pandemic-related data,and crawled the auto-related social media discussion data on Sina Weibo,Auto Home and BITAUTO.For market demand prediction in complex environment,this dissertation proposes a Boosting prediction method based on machine learning and compares it with traditional and machine learning methods such as historical average,SARIMA,SVM and ANN to verify the effectiveness of the proposed method.It is also found that the prediction effect of various methods is greatly affected when COVID-19 occurred.Therefore,this dissertation analyzes the performance of each prediction method combined with pandemic data,and the results show that combining pandemic data does not effectively improve the model prediction effect.For this reason,this dissertation applies natural language processing technology to mine social media data so as to grasp the real consumer demand,and integrates the social media mining information with the Boosting method,which shows that the social media information can effectively capture the auto market demand.After integrating the social media information,the Boosting model can achieve accurate prediction of auto market sales demand in each time period(before the pandemic,early pandemic,and post pandemic).In addition,experimental comparison analysis is conducted for different testing time periods and different data fusion methods to further verify the robustness of the proposed method.Next,after mastering the accurate market demand,a study is conducted on the mixed-model two-sided assembly line balancing problem to analyze the optimal balancing of the assembly line.Assembly line is one of the most personnel-intensive and time-consuming part of automobile production.The problem of finding its optimal balance is an important link in lean intelligent logistics.Mixed-model production is the mainstream mode for manufacturers at present.In this dissertation,we investigate the mixed-model two-sided assembly line balancing problem.Most of the existing studies have used heuristic algorithms to obtain approximate solutions,and there are few exact solution algorithms that can solve large-sized instances.Therefore,an exact solution algorithm based on combinatorial Benders decomposition is proposed to achieve the optimal solution including large-sized instances.In this dissertation,the problem is divided into a master problem,which is used to determine the distribution scheme of the task in each station,and a subproblem,which verifies the feasibility of the scheme provided by the master problem.For the infeasible case,this dissertation proposes a sequence-based enumerative search method to generate the cut set of the combinatorial Benders decomposition algorithm,including various search strategies such as Triangle Search,Cross Search,Snake Search and Inactive Task Search,and a pruning strategy Subset Pruning.Furthermore,the effectiveness and accuracy of the methods are verified by comparative analysis with mainstream heuristic and exact solution algorithms on small-sized and large-sized classical benchmark instances as well as different cycle times.Moreover,based on market demand and optimal balance information of assembly line,this dissertation focuses on the pickup problem of purchasing logistics.Most automobile manufacturers use the way of milk run to carry out parts purchasing currently.In order to further improve the leanness of logistics,this dissertation takes ProgressLane,a lean inventory approach,into consideration.With this approach,after the parts are delivered to the manufacturer from the supplier,the Progress-Lane is used to split the progress,buffer the difference between internal and external logistics progress,and realize the high-frequency small batch pickup of in-plant parts.This dissertation develops a model for the purchasing logistics problem integrating Progress-Lane.Meanwhile,to further verify the validity of the model,a zero-inventory comparison model is constructed and a genetic algorithm-based solution method is designed for each of the two models.In addition,this dissertation constructs instances of different scales based on real production data,and verifies the validity of the pickup model integrating Progress-Lane and the solution algorithm by comparing with the zero-inventory model.At last,after the parts arrive at the host plant,the research of in-plant intelligent distribution is carried out.On account of the trend of unmanned and decentralized intelligent manufacturing,this dissertation constructs a mathematical model of unmanned intelligent distribution to meet the production requirements such as decentralized inventory area,unmanned robotic distribution,and shareable equipment.Also,the model is rigorously proved to be a NP-hard problem.In view of the complexity of the problem,this dissertation proposes a heuristic algorithm based on greedy search to visualize the research problem,and constructs different scales of instances based on actual production scenarios to verify the effectiveness of the proposed model and method through experiments.Through experimental analysis,it is found that robotic delivery has good advantages,while robotic delivery based on milk-run delivery also has good practical value,which can meet relevant practical production scenarios.In addition,the maximum inventory along the assembly line is related to the robot delivery as well as the physical operation of the assembler,which should be determined scientifically when designing the production logistics.The corresponding inventory along the assembly line can be set appropriately within the allowed range of the physical area of the assembly plant.On the one hand,the above research expands the theoretical research in the field of parts logistics,on the other hand,the research intention of this paper is mostly inspired by the actual production of vehicle manufacturers,and the relevant research results have wide practical application scenarios,which can provide an important practical reference for manufactures to cope with the impact of COVID-19 pandemic,create lean intelligent parts logistics system,and carry out intelligent manufacturing upgrade and cost reduction. |
PDF Full Text Request