An Adaptive Scheduling Method Based On Manufacturing Cluster System Cognition

The manufacturing of aerospace products is a typical mixed flow production mode of single and small batch,which has poor repeatability of manufacturing process.The production process is easily affected by uncertainties and cannot follow the pre-defined benchmark plan.It relies heavily on manual off-line adjustment and planning.Therefore,how to build a new manufacturing system that can adapt to the highly coupled,dynamic and uncertain production environment and properly handle the contradictions and conflicts between workers,manufacturing resources and processing process is an urgent need of aerospace enterprises.Aiming at the scheduling requirements of aerospace products in dynamic uncertain environment,combining the thought of cluster system,and using the cognitive law of human being for reference,this paper studied on how to give cognitive ability to manufacturing cluster system to realize adaptive scheduling.The cognitive process technology system from the real-time state of manufacturing resources to the top level decision behavior is established to make the production system perfect its structure and function and enhance its adaptability to the environment.The main works of this paper are as follows:1.The realization principle of adaptive scheduling based on cognition of manufacturing cluster system is proposed.Grasping the state and its changes from different levels of production system,judging the impact of the changes and making a response resolution is the core issue to be solved in the cognition of manufacturing cluster system.By analyzing the characteristics and requirements of manufacturing cluster system and combining the cognitive law of human problem solving.This paper maps several steps of human problem solving cognitive to several levels of manufacturing cluster system cognition:defining to state cognition of manufacturing cluster system,searching for solution goals and paths to manufacturing situation cognition,generating and selecting the appropriate decision scheme to self-adaptive decision cognition.The cognitive architecture from the manufacturing cluster system state cognition to the manufacturing situation cognition,and then to the adaptive scheduling decision cognition is proposed.It includes cognitive goal,cognitive content,realization mechanism and so on,so as to overcome the influence of uncertain factors in the actual production process on the scheduling system.2.The realization method of manufacturing cluster system state cognition is proposed.In view of the time-varying of manufacturing resource state and the manufacturing resource state,and the nonlinear relation between the different levels of manufacturing cluster system combined state.In this paper,the multi-layer state topology and state transition mechanism of manufacturing resources in time series and combinational logic are established,and the manufacturing resource automata model,manufacturing unit automata model and manufacturing cluster automata model are established based on the hybrid automata method.The different manufacturing state transitions and their influence relations in the manufacturing cluster system are described.3.The realization method of manufacturing situational cognition is proposed.On the basis of obtaining the individual states of manufacturing resources,it is necessary to accurately judge the manufacturing situation formed by the combination of various processing factors,and realize the cognition of manufacturing situation.In view of the complexity of individual states and their combinations,and there is no linear logic relationship between individual states and manufacturing situations.A Bayesian network based manufacturing situation cognition method is proposed.The manufacturing situation is determined through the maximum posterior probability reasoning by learning the causality between individual states and manufacturing situation in historical production data.According to different cognitive levels,the corresponding learning algorithms are designed and implemented.There is no linear logic relationship between individual states and manufacturing unit situations,leading to the difficulty to directly map individual states to manufacturing unit situations.This paper proposes a multi-state fusion method for manufacturing unit situation cognition based on Bayesian Networks.In view of the non-independence of manufacturing unit situations,the relationship between manufacturing unit situations and manufacturing cluster system situations is complex and nonlinear.This paper proposes a method of situation cognition in manufacturing cluster system based on state entropy weighted naive Bayes.4.The realization method of self-adaptive scheduling decision cognition is proposed.After obtaining the manufacturing situation of the current manufacturing cluster system,and referring to the cognitive process of generating schemes and scheme optimization in human problem solving,a method of rescheduling scheme generation for manufacturing cluster system is proposed.Firstly,the adaptive index system of manufacturing cluster system for evaluation is constructed to judge the possible anomalies.Then,an anomaly classification method and an anomaly range discrimination method are established to determine the decision content of rescheduling.Finally,according to the hierarchical structure of manufacturing cluster system,combining the global searching ability of MCTS algorithm and the local optimization ability of MEO algorithm,four evaluation indexes,processing cost,processing quality,processing time and delay penalty,are taken as evaluation functions.A MCTS-EO algorithm for adaptive scheduling based on manufacturing situations in cluster system is designed to generate decision sets of manufacturing units for process processing and decision sets of manufacturing cluster system adapting to changes in situation sets of manufacturing units.5.Prototype system Design and Research results VerificationTaken the manufacturing system adaptive capacity of the mode of single small batch mixed flow production requirement as the background,developed the adaptive scheduling system based on cluster systems cognition(ASS-MCSC).Taking an aerospace box production workshop as an example,the theory,method and algorithm proposed in this paper are empirically studied and applied.The results show that the proposed method provides a new idea for the dynamic scheduling problem in actual aerospace production.