Health Evaluation And Safety Monitoring For Intelligent Elevators Based On Machine Learning Approaches

With social progress and the rapid development of modern cities,urban population is increasing day by day.High-rise buildings,such as large shopping malls,office buildings,high-rise residential areas and so on,have become an important means to alleviate the shortage of land in modern cities.As a vehicle connecting high-rise buildings vertically,elevator plays an extremely important role in daily life.However,despite the gradual improvement of elevator manufacturing and diagnosis technology,the number of elevator accidents is still on the rise with the continuous expansion of the total number of elevators and the increasingly frequent use of elevators.Therefore,how to ensure the safe and reliable operation of the elevator has become the focus of today’s society.At present,the main way to ensure elevator security is periodic maintenance,but the reliability is not enough.Focusing on the goal of ensuring the security of the elevator,which is a set of machinery,electrical,control as one of the complex special equipment,this paper researches the health evaluation of the normal operation of the elevator and the typical fault diagnosis of the abnormal operation of the elevator through data analysis by machine learning approaches.It realizes the timely evaluation of the health degree of the elevator and the accurate detection of the fault,and ensures the safety and reliability of the elevator operation from multiple levels.The specific research contents are as follows:(1)Traditional elevator operation performance evaluation methods are mostly based on objective indicators such as hardware performance and lack of real-time analysis and targeted feature extraction for operation data.It is difficult to give a real-time and effective health evaluation of elevator.Therefore,an evaluation method of elevator running health based on car acceleration signal analysis is proposed.This method deeply analyzes the main signal reflecting the elevator running quality,namely the car acceleration data.Through noise reduction and de-trend processing,the effective vibration signal of elevator is extracted.Then,the vibration signals are divided into time windows to calculate multiple evaluation indexes.Finally,the comprehensive score is obtained by bringing in the preset evaluation function,and the real-time and accurate evaluation of the health of the elevator is given,so as to provide a basis for the follow-up maintenance and protection of the elevator.(2)The vibration in elevator operation is closely related to the comfort and safety of passengers.At present,the detection of abnormal vibration of elevator lacks the specific cause of abnormal vibration diagnosis.Therefore,an elevator abnormal vibration diagnosis method based on multi-channel one-dimensional convolutional neural network is proposed.This method makes full use of the elevator vibration information.Firstly,the vibration signals are decomposed into multiple intrinsic mode function signals by empirical mode decomposition,and the abnormal vibration features of multiple angles are obtained.Then,the multi-channel one-dimensional convolutional neural network is constructed to carry out the multi-channel signal feature fusion and realize the accurate classification of various elevator abnormal vibrations.(3)The damage of elevator traction wire rope is a common and serious hidden danger.At present,the defect detection method of elevator traction wire rope damage generally has high complexity and labor cost.Therefore,a wear detection method for traction wire rope of elevator is proposed,which makes full use of the image texture features of traction wire rope.Firstly,the interference factors in the process of image acquisition are removed by means of grayscale and denoising pretreatment.Then,combining edge detection and template matching,the wear area of drag wire rope is determined according to the similarity measurement criterion.Finally,the image is finely segmented and the wear rates of different regions are calculated.This method can reduce the complexity of detection and detect the change of abrasion degree of elevator wire rope conveniently and intuitively.It can provide valuable reference for elevator maintenance.