Research On Elevator Failure Prediction Based On Knowledge Graph

With the increasing number of high-rise buildings,elevators have become common equipment for people’s production and life.However,due to the long-term and large-scale use of elevators,some potential safety hazards have also followed.When elevators are used,there are dangerous situations such as slip phenomenon of ladder,collapsing to the bottom of pit,and trapping in the elevator,so it is necessary to monitor the elevator running status data and predict elevator failures.At present,the amount of elevator-related monitoring data is gradually increasing,and the data processing model is becoming more and more complex.However,the traditional existing failure prediction methods have certain limitations,and the accuracy is difficult to improve.Aiming at the problems of repeated data,weak data relationship and slow update of elevator data in elevator operation system,elevator failure prediction system urgently needs to use new methods to avoid data deviation of elevator failure information and improve data storage efficiency and extraction process,so as to improve the performance of elevator failure prediction.This paper introduces the research background of elevator development and safety early warning at home and abroad,as well as the research status of elevator failure prediction technology at home and abroad,investigates the elevator failure prediction methods proposed by relevant scholars,compares the advantages and disadvantages of current algorithms,studies and judges the urgency of failure prediction in elevator safety,analyzes and summarizes the characteristics of elevator failure data,at the same time,combined with the structural semantic characteristics of knowledge graph and the characteristics that can improve the planning,data tracing and maintenance management of elevator equipment in the field of elevator equipment safety early warning,the elevator data and failure prediction problems such as repeated data,weak data relationship and slow update of elevator data in elevator operation system need to be improved are studied,The elevator failure prediction platform is built.The main research contents and contributions of this paper are as follows:1.Aiming at the problems of weak data relationship and repeated data of elevator,a method of elevator failure prediction based on knowledge graph combined with rule reasoning and knowledge graph embedding and completion is proposed in this paper.The elevator failure prediction is transformed into the task of embedding and complementing the knowledge graph,and the rule reasoning method is combined to construct the elevator failure prediction model,which mainly cleans and integrates the original data of the elevator failure,and the vibration value of the elevator traction machine collected by the sensor.Car vibration value,humidity,temperature,load capacity and several main attribute parameters are analyzed and stipulated to transform into structured data,combined with the ontology layer to form a knowledge graph;after the knowledge graph is constructed,the translation model is used to achieve the low triad dimensional vector representation,and finally combined with the Proj E（Embedding Projection）model to realize the task of complementing the knowledge graph,so as to obtain the elevator failure probability ranking obtained according to the parameter prediction.2.Aiming at the limitation of the above methods that are easier to produce fuzzy judgments,a solution is proposed to mine and predict elevator operating state data by using knowledge graphs combined with LSTM network algorithms.The Proj E model uses the ranking task to perform the failure prediction ranking.The result of the ranking is relatively coarse,and it is easier to produce fuzzy judgments in the prediction of elevator failure.In view of the deficiencies and limitations of the above models,and the complex characteristics of the relationship between the attributes and parameters of the elevator.In this paper,the knowledge graph is combined with the long-term and short-term memory network to further build a fine-grained elevator failure prediction model.After the knowledge graph triples have completed the low-dimensional dense vector embedding,the elevator failure prediction training and testing are carried out through LSTM network.During this period,the particle swarm algorithm is used to replace the traditional stochastic gradient descent process.The particles continuously change their positions and speeds.In order to obtain the optimal solution,and use it for the prediction of elevator failure.3.According to the proposed elevator failure prediction algorithm,this paper uses the frontend and back-end architecture to build an elevator failure prediction platform.The visualization of elevator failure prediction is realized through graphical components such as maps and charts.According to the research content of elevator failure prediction algorithm and model,with laravel framework as the back end,Vue framework as the front end,My SQL as the database to store common data,and neo4 j as the graph database to store knowledge graph.Through the linkage of different graphical modules,the development of elevator operation failure prediction platform under the front and rear end separation mode is completed.More intuitively and vividly show the elevator failure prediction in the form of maps and charts,and realize the transformation of failure prediction algorithm from experimental stage to application stage,which can provide help and reference for the establishment of elevator safety emergency mechanism and long-term maintenance mechanism.