| Elevator is a kind of necessary transportation equipment in the high quality development and production and life of modern cities,the number of elevators increases year by year,and casualties occur from time to time,which has aroused widespread concern of the society.At present,a wealth of research results have been achieved in the safety evaluation of traction elevator equipment at home and abroad.Based on the system engineering theory of “manmachine-environment”,this paper establishes the safety evaluation system of traction elevator through the combination weighting method of game theory and grey relational analysis(GRA)method,a traction elevator safety evaluation model based on convolutional neural network(CNN)was constructed,and compared with different models,the prototype design of the traction elevator evaluation system based on CNN was carried out in Axure RP software.The main research contents include:(1)Analyze the risk factors affecting traction elevator safety.Based on the elevator system structure,the common elevator accident types are summarized,and combined with the relevant technical specifications,the elevator safety risk factors are analyzed from three aspects:personnel,equipment and operating environment(2)Calculate index weight and construct evaluation index system.The construction principle of the indicator system is clearly defined,the elevator system safety is divided into 5evaluation levels,and 26 evaluation indicators are preliminarily selected.According to the game theory combination weighting method to determine the weight of indicators,grey correlation analysis(GRA)method is used to obtain the grey correlation degree of each indicator,and based on this,20 evaluation indicators are selected,and finally build the elevator safety evaluation index system.(3)A safety evaluation model of traction elevator based on convolutional neural network is established.Jupyter Notebook was used to build the network model,and the evaluation performance of the model was analyzed according to the confusion matrix,accuracy and loss function curve,the accuracy of the saved training model was 99.71%,and the accuracy of the test model was 99.56%.(4)Two kinds of neural network models,BP and GA-BP,were established as the control,and the accuracy of the models after testing were 97.8% and 99.21% respectively,the results show that the safety evaluation method of traction elevator based on CNN has certain advantages,and provides a new idea for elevator safety evaluation.Finally,a prototype of the evaluation system is designed in Axure RP software,which can make software developers understand the development requirements of the system more intuitively. |
