Functional Safety System Of Elevator Brake Real-time Status Monitoring And Prognostics Based On Unsupervised Deep Transfer Learning

With the continuous development of modernization,the safety of elevator systems has received further attention.As an important part of elevator security,elevator brakes are subject to effective physical parameter monitoring and prognostics.Based on the characteristic physical parameters that affect the operation of elevator brakes,this paper designs and builds a functional safety system based on unsupervised deep transfer learning.The system can monitor elevator brake temperature,noise,electrical signal,encoder signal,micro switch signal in real time and realize random fault judgment and prognostics of brake failure.The main work and results of this article are as follows:(1)The functional design requirements and reliability requirements of the elevator brake monitoring system is analyzed.According to the functional and reliability design requirements,the monitoring system is divided into three parts: sensor subsystem,logic subsystem,and final component subsystem.The overall frame design and modular design of the hardware system are completed.Finally,according to the hardware random failure analysis method which is proposed by IEC61508,the from-bottomto-top system reliability analysis and verification of the hardware system is carried out,which proves that the system meets the design requirements of second-level hardware safety integrity so as to meet the functional safety requirements.(2)An unsupervised deep transfer learning algorithm for fault prediction of elevator brake braking force decline is presented.The algorithm extracts the features of the original data with the help of the characteristics of long short-term memory encoder-decoder to detect outliers.Then remaining useful life is predicted by artificial neural networks based on the feature sequence.During transfer learning period,the maximum mean discrepancy error method is used to achieve the alignment of real data and simulated data in the feature space.The experiment results show that by taking the elevator brake clearance,braking noise,braking distance and acceleration as input,and combining with the simulation data for training,the algorithm can estimate the mean square error of the remaining life cycle of the elevator brake under a real environment of only 0.0016.It improves the accuracy of elevator brake failure prediction under real work and helps to improve the safety of special equipment such as elevators.(3)According to the functional requirements and reliability requirements of elevator brake monitoring,the software system is modularly designed.The physical parameters and calculation results collected by the embedded system are summarized on the cloud server,and the feature database is updated to provide new data guarantee for precise warning and follow-up research of elevator brakes.Finally,through the method of fault insertion test,the reliability of the monitoring system is verified.(4)For the test of elevator brake monitoring system,a set of elevator brake test platform and experiment plan was designed and verified.The platform simulates and accelerates the brake performance degradation process of the elevator by setting the torque loading,triggering speed and the test interval.During the experiment,the monitoring system conducted data collection and fault warning of multiple sets of elevator brakes throughout their life cycle,and finally achieved excellent prediction results.In the end,the real-time and reliability of the monitoring system were verified by the method of random fault insertion of the elevator brake.