Business Process Models Deep Generation For Smart Manufacturing

Germany,Britain and France have successively launched strategic plans for industry "4.0",industry "2050" and "Future Industry".The State Council of the country also announced the "Made in China 2025" plan in May 2015.It reflects the importance attached by various countries to industrial development.For the development of industry,the first thing is to develop industrial intelligence,that is,let machines help enterprises to complete complex tasks,reduce enterprise costs,and improve the quality of goods and enterprise services through artificial intelligence,deep learning and other methods.The core principles of industrial intelligence.The development of industrial intelligence is of great significance to the organization,management,transformation and optimization of enterprises.However,with the increasing complexity of the current industrial production environment,the industrial process model often faces the problem of being out of touch with reality in actual production.However,it takes a lot of time and money to construct an industrial process model artificially,and the process mining method will face the limitation that the process mining method cannot be used due to the lack of logs.Therefore,the automatic generation of industrial process models is the current demand of current industrial production,an important direction of industrial process development,and an enabling technology to drive high-efficiency operation of industrial production.In order to automatically generate industrial process models,based on the existing deep learning and natural language processing technology foundation,this thesis proposes a novel method for deep automatic generation of industrial process models from text descriptions of industrial process models.The method mainly includes three aspects:(1)Active entity recognition method based on industrial process text description.The method is based on the existing named entity method,constructs an active entity recognition model based on industrial process text description,aiming at extracting key information from industrial process text description;(2)Deep unsupervised retrieval of active entity hierarchy based on ordered neural network structure.Extend the language model to obtain the vector representation of active entities,and use ordered neural network(ON-LSTM)to unsupervised discover the potential hierarchical structure between active entities in the textual description of industrial processes;(3)The hierarchical structure between active entities is visualized and Automatic conversion to industrial process models.Firstly,the hierarchical structure of active entities is visualized as a hierarchical structure tree,the accuracy of the method proposed in this thesis is measured by the hierarchical structure tree,and the hierarchical structure tree is automatically converted into an industrial process model by using the principle of hierarchy depth.In this thesis,the feasibility and effectiveness of the method proposed in this thesis are verified by the manually acquired and annotated SAP industrial data.