Design And Implementation Of Key Technologies Of Intelligent Inspection System In Smart Pipe Gallery

With continuous evolution of urbanization in our country,underground pipelines such as electricity,gas,water supply and communications pipelines which are tightly related to all aspects of people’s daily life,are becoming more and more complicated.Problems such as frequent pipeline accidents and low efficiency of accident location directly affect the quality of the urban infrastructure services.The pipe gallery can be described as the "underground life network" and"supply network" of the city.As an important routine for the operation and maintenance of the pipe gallery,inspection is a basic work to ensure the normal operation of equipment and production safety,and it is responsible for detecting hidden dangers in advance and responding in a timely manner.With the development of artificial intelligence,machine learning and other technologies have brought new opportunities to predictive inspections.Image Recognition technology,Prognostics and Health Management(PHM)technology,Resource Scheduling and Path Planning technology etc.are key technologies of inspection system in smart pipe gallery,which play an important role in improving inspection efficiency.The main content of this thesis is to optimize the preparation and analysis work using PHM and path planning technologies before generating the inspection path based on the existing smart pipe gallery cloud platform.The main contents of this thesis are described as follows:(1)Design and implement the Remaining Useful Life(RUL)prediction function using Gradient Boosting Decision Tree(GBDT)algorithm and Long Short-Term Memory(LSTM)algorithm.In view of the excellent table data processing ability of GBDT,the LightGBM algorithm which is an improved algorithm of GBDT in engineering is selected as the baseline model to solve RUL prediction problem.Noting that the equipment data of smart pipe gallery is collected in time dimension,and GBDT does not have the ability of capturing features in time series.Therefore,this thesis proposes an enhanced GBDT method using LSTM,which is good at capturing time-related features,to improve the accuracy of the model effectively.In order to test the performance of GBDT-LSTM model,this thesis compares this model with other classical algorithms,such as support vector regression and GBDT algorithm.The result is that a set of evaluation indicators of the model perform better.(2)Design and implement a path planning method considering RUL of checkpoint equipment.The first step is to convert the predicted RUL value into equipment priority using a proposed standard.Then establish proper mathematical model and objective function according to actual inspection situation in smart pipe gallery.Through comparison with other algorithm,the genetic algorithm with good universality and robustness is selected as the basic algorithm.This thesis proposes an improved genetic algorithm using three optimization strategies for its efficiency problem and premature convergence:population initialization strategy using greedy algorithm,elite preserving strategy combined with roulette wheel selection strategy,one-way mutation strategy with mixed transposition and inversion mutation.This thesis designs and implements a path planning method based on the remaining useful life of the inspection equipment using the improved genetic algorithm.The simulation test and system realization prove that the path planning algorithm based on the remaining useful life of the inspection equipment can effectively reduce the inspection workload and realize the predictive inspection.(3)According to the actual requirements of the smart pipe gallery cloud platform,design and implement the intelligent inspection system.This thesis analyzes the requirements of the intelligent inspection system and gives the system architechure and technical solutions,realizes the remaining useful life prediction module and path planning module of the system,and tests and demonstrates the functions of the intelligent inspection system.With a series of functions like planning inspection path automatically,the intelligent inspection system realizes automatic inspection and targeted high-efficiency inspection without manual tracing.