Research And Application Of Intelligent Decision Support System For No-dig Mud Based On Detection While Drilling

At the beginning of 2022,in the government work report,the list of major projects and the first quarter housing construction work conference,all localities made it clear that this year’s urban pipe network transformation and construction construction drawings,and this field of people’s livelihood has also become the focus of concentrated construction of infrastructure projects in the first quarter of this year.The central economic work conference held at the end of last year proposed that during the 14 th Five Year Plan period,pipeline transformation and construction must be taken as an important infrastructure project.Subsequently,the state intensively issued a number of policies related to the construction of underground pipe network and water conservancy projects.With the "warm wind" blowing from the policy level,the agency predicts that the pipeline investment scale may exceed 1.4 trillion yuan during the 14 th Five Year Plan period.Under the environment of new environmental protection regulations and favorable policies,China’s no-dig industry is developing rapidly,but the corresponding supporting facilities of no-dig projects are difficult to meet the construction requirements,especially the no-dig mud technology.Due to its own characteristics,the construction area of no-dig engineering is often concentrated in buildings in the main urban area or river crossing section.It is difficult to carry out conventional engineering survey in such sections,conventional coring cannot be carried out,and the formation lithology is unknown,resulting in the "dark" state of no-dig construction.Moreover,due to the late start of no-dig engineering in China,the weak accumulation of talents and the shortage of mud engineers in the field of no-dig,it is difficult to determine the mud scheme matching with the formation lithology.No-dig mud scheme with poor compatibility with the stratum will lead to many types of construction accidents,such as the low density of the mud when driving into silt stratum,the difficulty in balancing the stratum stress,resulting in the instability of the hole wall,the insufficient lubrication of the mud driving into gravel stratum leads to the low driving efficiency and the delay of the construction period,and the insufficient inhibition of the mud driving into water sensitive stratum leads to the expansion and instability of the hole wall.Under the condition that there is no full-time drilling engineer to determine the formation lithology,and there is no full-time drilling engineer to solve the problem.To solve the above problems,this thesis combines detection while drilling technology,machine learning data mining and mud database to compile a set of no-dig mud intelligent decision-making system.,The formation lithology can be determined and the no-dig mud scheme can be decided only by inputting the no-dig drilling power parameters,mud rheology data or formation geotechnical parameters.Firstly,the experience of several survey and construction units in Shanghai area is summarized,and the typical no-dig strata in Shanghai area are classified to form the engineering geological zoning in Shanghai area.Based on the guidance and suggestions of experts in the field of mud,the corresponding mud performance indicators are proposed according to the characteristics of the stratum encountered by trenchless drilling and the process characteristics of the project,and the mud scheme is customized according to the mud performance indicators.Compile mud expert knowledge base with Visual C++and Access,and input more than 700 mud schemes and mud properties.Establish a project case information database to record the mud information of completed trenchless works,and provide mud reference for subsequent no-dig works with similar geographical location and construction technology.Secondly,a set of detection system while drilling suitable for no-dig is designed to collect the dynamic parameter data of near bit and rheological property data of flowback mud generated when no-dig enters various types of formations,providing data support for the establishment of subsequent formation identification model.Near bit torque,drill pipe axial force,rotation speed and mud pressure can be detected by near bit dynamic parameter detection module.The density,apparent viscosity and plastic viscosity of flowback mud can be detected by the rheological parameter detection module of flowback mud.The dynamic parameters near the drill bit can be transmitted in real time by electromagnetic wave wireless transmission device,which is convenient for data acquisition and subsequent model application in the project site.After the MWD data is obtained,preprocessing is performed for the formation identification model construction.The random forest algorithm is applied to train the dynamic parameters of the near bit to form a stratum recognition model.The recognition results are reduced by PCA to achieve 3D visualization.The rheological parameters of flowback mud are classified through the serial application of KNN-SVM algorithm,and the formation identification model of weak supervision machine learning is established.Among them,KNN algorithm is used to transfer mud data labels,providing a data basis for SVM formation lithology prediction.Through the actual measurement of no-dig construction in Pudong,Shanghai,the validity of the stratum identification model is verified.The accuracy of the random forest stratum identification model is 92%,and the accuracy of the SVM stratum identification model is 96%,which proves that both types of models can effectively identify the no-dig stratum in Pudong,Shanghai.The establishment of the two types of stratum identification models provides important geological information for the selection of drilling tools and mud design in the no-dig reaming stage.Finally,the algorithm model was compiled into the no-dig mud intelligent decision-making system,and the "Formation Lithology Identification Model Based on Random Forest and Near Bit While Drilling Parameters" and "Formation Lithology Identification Model Based on Support Vector Machine and Mud Rheological Parameters" were compiled into the system model library,and "Formation Lithology Identification Model Based on Production Rules" and "Formation Lithology Identification Model Based on Engineering Geological Zoning" were created.In this way,only the rheological parameters of mud or the dynamic parameters near the bit can be input to determine the formation lithology,and only the geotechnical parameters or engineering coordinates can be input to determine the formation lithology when the parameters while drilling cannot be obtained.The inference engine of the no-dig mud intelligent decision-making system links the "mud expert knowledge base" according to the formation lithology discrimination results to decide the mud scheme.Web Api,the electronic map of Shanghai area,is compiled into the no-dig mud intelligent decision-making system to realize the map visualization of engineering information.To sum up,a series of theoretical and practical research work on the compilation of no-dig mud intelligent decision system has been carried out by this thesis.It is a typical interdisciplinary application.Relevant research results have been applied in engineering practice and achieved the expected assumption,providing technical and theoretical support for the follow-up No-dig informatization and intellectualization.In the future,the author will still work deeply in the field of no-dig,and intends to contribute his modest efforts to the cause of no-dig in China in the aspects of artificial intelligence,MEMS application and the preparation of relevant specifications of no-dig industry.