Elevator Health Assessment Based On Logistic Regression

As a vehicle for the last 50 meters of modern human activities,the elevator has become one of the important infrastructure closely related to the life of the urban residents.The existing management methods have played an important role in the maintenance of’elevator in our country in the past 10 years,but the rapid growth of the elevator has brought great challenges to the traditional management.The fault of the elevator is related to its structure and history.The real-time running state is a comprehensive expression of the historical running state.The historical running state reflects the health status of the elevator to a certain extent.The historical data include the operation data,faults and causes,maintenance records of the elevator.This project is based on historical data and real-time operation data.By introducing Al method,we optimize the elevator management method,explore new management mode,try to reduce the failure rate of the elevator,and guarantee the normal and orderly production and life.First,by analyzing the cause and design structure of the accident,combining with relevant design,maintenance and operation standards,we completed the rough extraction of elevator health characteristics.On the basis of the existing elevator data,through the variance threshold filtering method,eliminating the characteristic data of little fluctuation is small,low content of information entropy,greatly reduced the feature dimension;feature extraction characteristics respectively by logistic regression and SVM algorithm as the foundation,further screening characteristics;using the random forest algorithm,using different theory an algorithm of feature;combining the recursive feature selection method and evaluation algorithm,further screened for feature lift health evaluation.The evaluation features are integrated by artificial intelligence and machine learning algorithm.Then,delete,insert and other features into the mode method to fill the missing data;using one-hot algorithm,improved the data does not match the change of business environment;with the help of data standardization,optimization problems due to a decline in performance evaluation algorithm of data units are not unified cause;a combination of K-Means++ and SMOTE algorithm,a new data derived based on the existing data,solves the data imbalance problem.Finally,based on logistic regression theory,the small batch gradient descent algorithm,Adagrad algorithm and high-order linear model are introduced to optimize the algorithm efficiency and improve the accuracy of the evaluation model,and preliminarily completed the evaluation and early warning for elevator health.