Research And Application Of Knowledge Work Automation In The Process Industry

The production and operation of process industry,especially complex petrochemical process,relies heavily on the experience,methods and knowledge of experts and staff.In the industrial analysis report,McKinsey thinks that knowledge worker automation will become the subversive technology with the second economic benefit in the future.Knowledge automation is an advanced intelligent knowledge automation system including knowledge representation,knowledge generation,knowledge reasoning evolution and knowledge integration application.It not only includes the rules,reasoning and algorithms of traditional knowledge system,but also models the tacit knowledge,such as production and operation mode,which is difficult to express.Referring to the standardized knowledge representation form,the machine can easily identify and explain the knowledge warehouse,so that the computer can replace the knowledge worker to complete a large number of patterned and knowledgeable work,and further release the knowledge workers innovative productivity.In the process industry,the traditional mechanism reaction knowledge,such as the complex chemical reaction mechanism has been solidified into the physical production process.Therefore,this dissertation focuses on the knowledge automation process which relies heavily on the experience of experts.The purpose of this dissertation is to propose a formal knowledge representation framework for complex process industry process systems,and the strategy of knowledge storage,generation,reasoning and evaluation based on this framework,so as to solve the problems in process industry,such as the diversity of data types,the difficulty of knowledge mining,the big difference of unified description standards and the organic integration of data and knowledge;combined with the hierarchical structure of process industry process system(unit,device,system),mechanism model of production process,process flow diagram(PFD diagram),P&ID diagram(Piping&Instrument Diagram),the connection relationship among production operation units,equipment and units is analyzed;the method of modeling,reasoning operation and evaluation and analysis based on data and knowledge for the integration of logistics,energy flow and information flow(multi unit,multi device,multi process,etc.)is formed.In this dissertation,the knowledge automation of ethylene production in process industry is studied:(1)Due to the lack of corpus data in process industry,it is impossible to mine important feature knowledge by bottom-up pattern,and then construct knowledge ontology;similarly,it is very important to define the application scenarios of knowledge in process industry,and knowledge workers should confirm which knowledge is applied to which field and solve what kind of problems,but the content composition of knowledge gradually changes with time and complexity of working conditions It is also difficult to construct knowledge ontology from top to bottom and solve the problem completely.Therefore,this dissertation combines iso-15926 standard and p-graph theory to construct a knowledge level ontology,and designs a top-down and bottom-up iterative ontology updating strategy,which provides a framework for the formal standardization of equipment,operation,materials and related application knowledge in process industry.(2)Application scenarios should be set in advance for the instantiation of vertical domain knowledge in process industry,such as raw material scheduling for ethylene production,operation optimization of cracking furnace group,etc.Besides understanding the operation mechanism of production process and the characteristics of production data,more knowledge workers are required to participate in the design of application scenarios and knowledge modules.In this dissertation,aiming at the scheduling optimization problem of raw materials for ethylene production,based on the idea of hierarchical ontology,a superstructure model of raw material scheduling and allocation solutions is constructed.The hierarchical ontology and data are automatically mapped to knowledge rules,and the RDF(resource description)is generated.Finally,a case study of ethylene raw material distribution shows the specific form of knowledge of distribution solutions.(3)Because of the free expansion of RDF semantic data,its ability to express semantic information and no model constraints or constraints in the specific application calculation process,RDF is more and more used in industrial systems.However,knowledge workers are still needed to participate in problems such as material scheduling in factories,and semantic data has not been fully utilized to realize automatic modeling.Therefore,this dissertation proposes a method to construct RDF triplet data based on the atomic model and coupling model of devs(discrete event system specification).While standardizing the semantic knowledge model of data,taking advantage of the advantages of graph data structure storage,the RDF based maxial structure generation(RDF-based maxial structure generation)algorithm is proposed.The difference between rdfmsg and traditional MSG is that it does not delete the original structure first and then reorganize it.Instead,it only performs one traversal of RDF data,which greatly simplifies the algorithm process and reduces the execution complexity of the algorithm.(4)Due to the large differences in process,technology and scale of each ethylene production plant,the cost,output and carbon emission of the final raw material scheduling solutions are very different.The production decision-maker will refer to the actual working conditions and production objectives when selecting the raw material scheduling solutions,but it is difficult to take into account the complexity and completeness.Therefore,the improvement of the case base is a step-by-step iterative process.At the same time,the case base may not contain a solution that meets the requirements,so knowledge workers need to carry out professional reasoning to generate the required solution.This dissertation proposes an innovative case-based reasoning system,which forms a complete process of case representation,case base construction,case retrieval and case-based reasoning evolution.The knowledge base can provide a better reference for sustainable development.(5)The reasonable selection of material scheduling and allocation solutions is of great importance to the production cost and environment.Similarly,factors such as production scale and technology also affect the selection of material scheduling solution.Therefore,the evaluation and selection strategy of case knowledge by domain experts is also an important content of knowledge automation.In this dissertation,a case evaluation and analysis method integrating deviation grouping and data envelopment analysis(DEA)is proposed,so that the decision-making process such as case selection can achieve a certain degree of autonomous decision-making.The set of Top-N optimal scheduling solutions is generated based on the superstructure model constructed in the previous paper,and the ISA-95 standard and DEA evaluation method are integrated to analyze the material consumption and product output,and the solutions selection strategy is generated from the input-output perspective.In addition,integrating the above chapters,we can update the hierarchical ontology and instantiation knowledge proposed in this dissertation,and form a complete system for the research work of process industry knowledge automation.