Bottleneck Identification And Drift Prediction Of Discrete Manufacturing System In An Uncertain Environment

Bottleneck machine is the key to restrict the productivity of a discrete manufacturing system,and identifying bottleneck machine is one of the important research directions of a discrete manufacturing system monitoring.The uncertainty of a discrete manufacturing system,the flexibility of process routes,the limitation of production resources and the diversity of processed products make the bottleneck identification in a discrete manufacturing system very complicated.Influenced by different system objects and operational objectives,the definition of bottleneck depends on specific research problems.In an uncertain environment,the spatial position of the bottleneck may also shift.On the basis of identifying bottlenecks,it is more practical to study the dynamic changes of bottlenecks with time and predict them in time.For the bottleneck identification and prediction in a discrete manufacturing system,the bottleneck identification and prediction methods under several typical working conditions are studied as follows:(1)Aiming at a large-scale discrete manufacturing system,it is considered that the amount of production data increases greatly with the increase of system scale.A directed weighted production system network model is established by using complex network,and then the manufacturing units in the system are screened by using the topological characteristics of complex network to obtain candidate bottleneck machine groups,so as to reduce the amount of data collected and analyzed for bottleneck identification.Aiming at the candidate bottleneck machines,the bottleneck decision matrix is constructed according to the machine utilization rate,the average machine activity rate and the total energy consumption of a single machine.Define the degree of separation between candidate bottleneck machines,get the optimal weight of each attribute based on the idea of maximizing the degree of separation,and get the comprehensive bottleneck value through the optimal weight and attribute value to judge the bottleneck machine.(2)Aiming at a discrete manufacturing system with fluctuating processing time,it is difficult to describe the bottleneck evaluation attributes of manufacturing units with definite values.The interval number is used to describe the machine utilization rate,the average machine activity rate and the total energy consumption of a single machine,and an interval bottleneck decision matrix is constructed.The optimal weight is determined by maximizing the degree of separation,and the overall value of the range bottleneck is calculated from the optimal weight and the values of the range attributes.Furthermore,considering that it is difficult to directly compare the comprehensive bottleneck values of interval type.The possibility matrix of bottleneck machines is defined.And convert the possibility matrix into a fuzzy consistent complementary judgment matrix to calculate the rating value of bottleneck machines.Finally,introduce the Fuzzy C-means(FCM)algorithm for clustering machines.And the bottleneck clusters of I,II and III levels are obtained,thus realizing the classification of the bottleneck machines.(3)Aiming at a discrete manufacturing system with transition probability,in the uncertain environment,the bottleneck position may shift in space,that is,bottleneck drift occurs.In order to predict the bottleneck drift,a bottleneck prediction method based on grey neural network is proposed.Considering the fluctuation of bottleneck index value,GM(2,1)model,which is more suitable for non-monotonic oscillation development sequence prediction than GM(1,1),is adopted,and the average weakening buffer operator is introduced,and the equal-dimension and new-information GM(2,1)model is constructed by combining the metabolism idea.At the same time,BP neural network is introduced to modify the prediction residual of GM(2,1),and a dynamic bottleneck prediction model based on grey neural network is constructed to obtain the predicted value of the bottleneck index of manufacturing units,which effectively realizes the drift prediction of the bottleneck position.