Research On Data-driven Key Performance Prediction And Optimization Method For Io T-based Manufacturing System

With the acceleration of economic globalization and the intensification of market competition,there are some brand-new changes in the manufacturing industry.For example,the manufacturing paradigm is transformed from manual,extensive and passive mode to transparent,lean and proactive mode.However,due to the lack of real-time monitor and accurate prediction method of key production performance,the upper-level production management system cannot get the real-time information of manufacturing execution processes,which may further lead to an adverse impact on the dynamic control of manufacturing processes and optimal allocation of manufacturing resources.With the rapid development of advanced information technologies such as cloud computing,industrial Internet of Thing(Io T),and cyber physical systems(CPS),the concept of the Io T-based manufacturing system(Io TMS)was proposed.In the Io TMS,many advanced management measures have been enabled,such as timely collection of manufacturing data,rapid response of production disturbance and accurate execution of production decisions.As a result,it is essential to upgrade the existing key performance prediction and manufacturing resources optimization method to a more efficient level.Under this background,this paper proposes a data-driven key performance prediction and optimization method for Io T-based manufacturing system.The discrete manufacturing shopfloor is taken as the application environment,the advanced manufacturing technologies,Internet of Things technologies,CPS technologies and information prediction technologies are combined in the proposed method.After the overall architecture of the method is presented,the core technologies such as real-time analysis,dynamic prediction and proactive optimization for Io MT are discussed in detail.The main contents of the thesis are concluded as follows:(1)Overall architecture of data-driven key performance prediction and optimization method for Io T-based manufacturing systemIn order to meet the challenges of dynamic performance prediction and optimal resources allocation in Io T-based manufacturing system,three related concepts,including CPS-enabled manufacturing system,data sensing technologies in Io T-based manufacturing system,and key performance indicators of manufacturing systems are introduced firstly.Then,the overall architecture and operational logic of the proposed method is presented.Finally,three supporting key technologies are proposed and briefly discussed.(2)Multi-source and real-time manufacturing data-driven key performance analysis model for Io T-based manufacturing systemIn order to meet the requirements of information integration and exception diagnosis among composite events of key performance and primitive events of manufacturing resources,a real-time key performance analysis model for Io T-based manufacturing system is proposed.Three core technologies of the model are discussed,namely event-driven active key performance sensing of manufacturing system,decision tree(DT)-based dynamic identification of key performance exception and fuzzy Bayesian network(FBN)-based production exception diagnosis method.Based on the application of the proposed method,the operational status and key performance of manufacturing system can be timely perceived and evaluated,which can provide information support for following dynamic performance prediction and proactive decision making.(3)Dynamic key performance prediction method for Internet of manufacturing things in a multi-state production environmentIn the current complex production system,the state of production environment always changes randomly and it is hard to describe the evolution mechanism.A dynamic key performance prediction method for Io T-based manufacturing system is proposed by combing historic manufacturing data,real-time production status and expert knowledge.Based on this,three critical production performance prediction models are discussed in detail,namely Dynamic Bayesian Network(DBN)-based manufacturing service capacity prediction,Hierarchical Timed Coloured Petri Net(HTCPN)-based work-in-progress(WIP)processing time prediction and Deep Learning(DL)-based production bottleneck prediction method.As a result,accurate information about key performance can be provided for proactive decisionmaking.(4)Dynamic prediction information-enabled proactive key performance optimization method for Io T-based manufacturing systemIn the existing process control and resource optimization strategies,it is hard to respond abnormal events timely and obtain optimal production decisions.In order to address these issues,a proactive key performance optimization method for Io T-based manufacturing system is proposed,and its components and work logic are discussed in detail.Then,two proactive optimization models are presented as examples,namely proactive optimization method of manufacturing tasks and proactive production material handling.The former model can adjust production task queues before production exceptions arrive,while the latter model can allocate the logistics tasks predictively before WIPs are processed.The Non-dominated Sorting Genetic Algorithm-II(NSGA-II)is used to obtain optimal decisions for the two models.The proposed method can provide an accurate,efficient and autonomous process for optimal configuration of manufacturing resources.(5)Simulation system development and case verificationBased on the above key enabling technologies,a simulation system of key performance prediction and optimization for Io T-based manufacturing system is established based on a final machine assembly shop-floor.Three key modules of the simulation system are discussed in detail to verify the effectiveness and feasibility of the proposed method.