Research On Dynamic Scheduling Of Intelligent Workshop Manufacturing Process Based On Digital Twin

In Digital Twin Job-shop(DTJ),the environment of the job-shop has changed greatly compared with the present manufacturing job-shop.The dimension and depth of manufacturing resource data are increasing.Manufacturing process data,which has not been fully integrated into the scheduling decision system in the research of job shop scheduling,can also be fully captured for the development of Internet of things technologies such as sensors and embedded system.With the increasing demand of digital twin job-shop scheduling,the existing scheduling method based on manufacturing resources can not meet the demand of digital twin job-shop scheduling accuracy.Therefore,how to use manufacturing process data to implement scheduling decision in digital twin job-shop is worthy of attention and has research significance.The research is based on the National Natural Science Foundation of China(NSFC)project "research on self-organizing scenario modeling and application of digital job-shop manufacturing behavior"(51865004).Based on the design of information perception and fusion platform of digital twin job-shop,the dynamic scheduling optimization model and algorithm of manufacturing scheduling are constructed,and a prototype system is developed according to the requirement of the enterprise.In the research,the improved models such as dynamic scheduling model,data collection and analysis model and dynamic scheduling knowledge model of digital twin shop based on edge computing,and the improved algorithms such as hybrid knowledge mining algorithm based on Decision Tree(DT)and gravity search optimization algorithm based on Differential Evolution(DE),play an active role in improving the scheduling efficiency of digital twin job-shop scheduling and improving the intelligent manufacturing level of production shop.The main research contents and innovation points are as follows.(1)Aiming at the problem of information perception and fusion in digital twin job-shop,the information perception and fusion platform architecture of digital twin job-shop for dynamic scheduling of manufacturing process is studied.This paper presents a scheme to collect job-shop manufacturing data by means of 3D imaging,(wireless)sensor,(wireless)sensor network and embedded system technology,and designs the interface protocol of job-shop equipment based on OPC UA and the method of data fusion at the bottom layer.(2)To solve the problem of resource information standardization in digital twin job-shop scheduling,the property,interface and function of five key factors in digital twin job-shop are analyzed which are product/component(material),equipment,person,production environment and production knowledge.Based on this,a dynamic scheduling model oriented digital twin job-shop is established,and its operation mechanism and logical mapping are studied.According to the manufacturing process of digital twin job-shop,a scheduling-oriented data acquisition and analysis model is proposed.The characterization and specification of manufacturing process data is implemented by mining scheduling rules and applying scheduling rules.The adaptive data acquisition strategy is adopted.The sampling frequency is adjusted according to the variance of the data obtained by the sensor during the sampling period.The Bartlett test is introduced to test whether the data set collected by the sensor in several cycles is from a population with equal variance.If the data set collected by the sensor has the same variance over several cycles,the sampling frequency will decrease;otherwise,the sampling frequency will increase.(3)Aiming at the problem of knowledge mining for scheduling rules in manufacturing process data,a dynamic scheduling knowledge model for digital twinning is proposed,and a hybrid knowledge mining algorithm based on decision tree is designed.The manufacturing process data generated during the implementation of the initial/pre-executed scheduling scheme are fused,analyzed and extracted to enter the knowledge model as training examples.Three Mining Algorithms,Decision Tree(DT),Random Forest(RT)and Radial Basis Function Neural Network(RBFNN),are started by the knowledge model to mine the scheduling knowledge,which will guide the scheduling optimization.The continuous evolution of the knowledge model,the improvement of scheduling knowledge decision-making level and the improvement of digital twin job-shop operation efficiency can be realized by the above-mentioned process cycle.(4)To solve the optimization problem of manufacturing process scheduling in digital twin job-shop,a dynamic scheduling model structure(Dynamic Scheduling model of Manufacturing Process in Digital Twin Job-shop based on Edge Computing,ECDTJ-DC)based on edge computing.To implement the integrated scheduling decision guided by the scheduling rules form manufacturing process data mining.Based on the investigation of more than 30 manufacturing enterprises,a gravity search optimal scheduling method based on differential evolution(DE-based Gravity Search Algorithm,DGSA)is proposed to solve the problem of blocking flow in manufacturing process with concentrated response.The goal of the algorithm is to minimize the Total Flow Time(TFT).The VNH operator is designed to initialize the heuristic population.The search efficiency is improved by the setting a new mode of acceleration,velocity and position of particles in gravitational search.The adaptive perturbation strategy based on hamming distance test is designed to improve the diversity of population,and the parameter setting method based on Average Relative Transient Percentage(ARTP)is designed.The experimental results show that this method has better scheduling performance,real-time performance and stability.(5)Set up a test environment in the laboratory for validation of relevant models,includes the validity of the proposed dynamic scheduling model based on edge computing(ECDTJ-DC),the data collection and analysis model based on ECDTJ-DC model and the dynamic scheduling knowledge model based on ECDTJ-DC model are tested.The test results show that the data acquisition platform is feasible,the three models run effectively,and the knowledge model can mine the scheduling knowledge rules effectively.At last,in view of the actual requirements of the intelligent transformation of a vehicle manufacturing enterprise and its job-shops in Guizhou,the job-shop scheduling and early warning module and the module with data twin application foreground are improved.The feasibility and validity of the proposed model and algorithm are verified,and it also provides support for the next step of constructing panoramic digital twin intelligent job-shop or even intelligent factory.