Study On Key Technologies Of Reconfigurable Shop Floor Scheduling For Customized Manufacturing

In the customized manufacturing,the changing type and volume of products,the constrains between quality and cost,and the sudden reduction in lead times make it difficult for traditional static shop floor scheduling method to cope with continuously changing production.The reconfigurable shop floor scheduling approach provides a new way of thinking for the dynamic resource reconfiguration to balance production efficiency and production flexibility.This paper takes the reconfigurable shop floor scheduling key technology for customized manufacturing as the research object,to cope with different production demands through selforganised resource reconfiguration,and production interruptions through adaptive dynamic optimisation scheduling.The research aims to establish a reconfigurable shop floor scheduling method with multiple indicators such as production time,production efficiency and production load,so as to achieve collaboration,forecasting,optimisation and control of production planning at the physical level and cyber level.The main original work is summarised in the following four areas.(1)The relationship between resource reconfiguration and scheduling in reconfigurable shop floor is analysed in the context of the customized manufacutring.Based on the characteristics of flow shop and job shop,the mathematical representation of the reconfigurable shop floor scheduling problem is established,which is used to propose the scheduling indicators the reconfigurable shop floor scheduling performance.A hybrid distributed control architecture based on Multi-agent theory is established to meet the distributed management demands of products and resources.Besides,the agent interaction behaviour modelling method for resource reconfiguration and scheduling is proposed.(2)For shop floor resource reconfiguration,A self-organising reconfiguration method that integrates ontology knowledge reasoning and agent negotiation mechanism is proposed.In order to achieve ambiguity-free semantic interoperation between products and resources in the reconfiguration scenario,ontology knowledge description of resource reconfiguration-related concepts is described with description logic,and a multi-level matching method for product and resource matching based on ontology mapping is proposed.Based on semantic rule reasoning and semantic query,the autonomous adjustment for shop floor capability is realized for two scenarios: resource abnormality inside the workshop and external production demand change.Additionally,The agent negotiation mechanism based on improved contract network protocol is proposed to ensure smooth resource reconfiguration.(3)A reinforcement learning-based adaptive scheduling optimization method is investigated with manufacturing units as the scheduling object.Based on the mathematical description of the reconfigurable shop floor scheduling problem,state characteristics,action spaces,payoffs and value functions of the semi-Markov Decision Process(SMDP)model for manufacturing task allocation are constructed.For the online solution of the SMDP model,an improved Q-learning algorithm based on linear function approximation,a policy-value learning algorithm based on convolutional neural network approximation are proposed to respectively ensure the solution speed of the small-scale state-space scheduling problem and the solution performance of the large-scale state-space scheduling problem.The proposed model training technology based on the virtual commission platform achieves improvement of model solving,and the periodic adjustment of the model parameters is achieved by constructing the adaptive retraining mechanism.(4)The middleware technology of reconfigurable shop floor management for continuous integration and deployment is proposed and implemented.Based on the concept of continuous integration and deployment,the service-oriented middleware structure is proposed,and the operation logic of continuous integration and deployment is realised by the collaboration of the middleware modules.To address the integration difficulty of hardware and software in the management system,the plug-and-produce based intelligent production edge adapter is designed to realise service-based middleware access to products and resources.A multigranularity and multi-view information modelling approach is adopted to carry out Automation ML information modelling of products and resources.Through the equivalence mapping of the Automation ML information model to methods and variable nodes in OPC-UA,a standard service encapsulation of the information model is realized,to meet the information extraction in different lifecycle stages of the reconfigurable shop floor.Finally,the feasibilities of the proposed methods are verified in the prototype platform of the reconfigurable shop floor management system.In summary,this paper presents theoretical methods and implementation techniques for reconfigurable shop floor scheduling under the scenarios of external production demand change and internal resource failure recovery.