| In order to realize the high-quality transformation from a "manufacturing power" to a "manufacturing power",China has clearly proposed to take intelligent manufacturing as the main focus of the transformation,upgrading and innovative development of the manufacturing industry,and through the integration and application of the new generation of information technology,to promote the manufacturing industry to move up the value chain and high-quality development.Intelligent workshop is the basic carrier of intelligent manufacturing mode.Under the dynamic and changeable manufacturing environment,it is the key to ensure the order period and production efficiency to independently and dynamically assign the production tasks in the intelligent workshop based on the data collection of the manufacturing process.Firstly,the characteristics of traditional production process and production data were analyzed,from which the data types needed to solve the problem of production task assignment were extracted,and real-time data of manufacturing process were collected through RFID,sensors,Zig Bee and other IOT technologies.The intelligent workshop production process is discretized into independent production events,and simple and abnormal events in the production process are generated based on the collected production process data.Secondly,according to the characteristics of the production tasks of autonomous dynamic points matching problem,the problem combined with hidden markov model(HMM),is proposed based on hidden markov model(HMM)production tasks dynamic allocation methods,training model,using the algorithm of Baum Welch-tau machine tool selection coefficient is put forward to improve the Viterbi algorithm and solving,implements the production task dynamically allocated for abnormal events.The feasibility of the method is verified by a case study.Thirdly,in view of the HMM observation independence assumptions and forecast results unreasonable problems,puts forward the improved maximum entropy markov model(D＿MEMM),introducing the characteristic function,and USES the GIS training algorithm parameters,the Viterbi algorithm transformed D＿MEMM algorithm,is established considering the logistics distance constraints between machine tool and production task dynamic allocation model.The rationality of this method compared with HMM model is verified by a case study.Finally,based on the same data set,four methods,HMM,D＿MEMM,bayesian rule prediction(BR)and regression analysis prediction(REG),were compared and analyzed,and then the influence of process number and data size on the performance of each method was analyzed.The results show that HMM and D＿MEMM have high accuracy and reliability in solving production task assignment,and D＿MEMM can still maintain good prediction ability in the case of small data scale. |
