CN-121981711-A - Elevator intelligent maintenance method and device based on DGConv and CBAM attention mechanism improvement

CN121981711ACN 121981711 ACN121981711 ACN 121981711ACN-121981711-A

Abstract

The invention discloses an intelligent maintenance method and device for an elevator based on DGConv and an improved CBAM attention mechanism, wherein continuous data and discrete data related to faults are firstly obtained based on an elevator main board, data length alignment is carried out on the continuous data through dynamic time warping, an elevator fault prediction data set is built after the continuous data and the discrete data are combined, a CNN-transducer mixed neural network of a cavity global convolution and the improved CBAM attention mechanism is then built and used for predicting occurrence probability of various faults at the next moment and training based on the elevator fault prediction data set, and finally fault maintenance priorities are distributed to different faults according to the fault probability, so that a targeted intelligent maintenance strategy is generated. The method provided by the invention has remarkable effect in the intelligent maintenance of the elevator, can greatly reduce the manpower and material resource expenditure of the traditional maintenance method, saves the maintenance time and optimizes the maintenance strategy.

Inventors

CHEN DONGDONG
CHEN LIN
LIANG LIHUA
LIU WANBING
WANG LIBIN
LANG YUE
FENG JINKUI
JIN XIAOHANG
LIN ZHENG
GUO XUFENG

Assignees

浙江省特种设备科学研究院
浙江工业大学
恒达富士电梯有限公司

Dates

Publication Date: 20260505
Application Date: 20260407

Claims (10)

1. An intelligent maintenance method for an elevator based on DGConv and an improved CBAM attention mechanism, the method comprising: (1) Based on the elevator main board, acquiring continuous data and discrete data related to faults, aligning the data length of the continuous data through dynamic time warping, and merging the continuous data with the discrete data to construct an elevator fault prediction data set; (2) The CNN-transducer mixed neural network of the cavity global convolution and the improved CBAM attention mechanism is constructed and used for predicting occurrence probability of various faults at the next moment and training based on an elevator fault prediction dataset, and the improved CBAM attention mechanism is that a global standard difference pooling path is added in a channel attention mechanism and a space attention mechanism to form a complementary statistical view with the existing information; (3) After the occurrence probability of the fault at the next moment is obtained, fault maintenance priorities are distributed to different faults according to the fault probability, and a targeted intelligent maintenance strategy is generated.
2. The intelligent maintenance method of an elevator based on DGConv and improved CBAM attention mechanism of claim 1, wherein in step (1) continuous data and discrete data are obtained for a period of time before and at the time of occurrence of an elevator fault, wherein the continuous data include fault type, fault duration, fault occurrence frequency, fault severity, leveling signal status, door lock status, drive system status and safety loop status, and the discrete data include voltage current data, running speed data and temperature data.
3. The intelligent elevator maintenance method based on DGConv and improving CBAM attention mechanisms according to claim 1, wherein in the step (1), after continuous data are aligned in data length, time domain indexes and frequency domain indexes are extracted, wherein the time domain indexes comprise mean value, root mean square value, peak-peak value, standard deviation, skewness, kurtosis, waveform index, peak value index and pulse index, and the frequency domain indexes comprise center of gravity frequency, mean square frequency, frequency variance, frequency spectrum kurtosis, side band analysis index and total harmonic distortion rate.
4. The intelligent maintenance method for the elevator based on DGConv and improved CBAM attention mechanisms is characterized in that in the step (2), an elevator fault prediction data set is used as a training set, the data is divided into two identical paths, one path sequentially passes through three groups of continuous cavity global convolution layers, an improved CBAM attention layer and a ReLU layer, the other path passes through a transducer neural network, feature level data fusion is achieved through the two paths of output through a cross attention mechanism, and finally occurrence probability of various faults at the next moment is obtained through a full connection layer and a Softmax activation function.
5. The intelligent maintenance method of an elevator based on DGConv and improved CBAM attention mechanisms according to claim 1, wherein in step (2), three paths are obtained in the channel attention mechanism by global average pooling, global maximum pooling and global standard deviation pooling respectively, and the three paths are multiplied by a shared multi-layer convolutional neural network channel by channel according to different coefficients respectively, so as to obtain an output result.
6. The intelligent maintenance method of an elevator based on DGConv and improving CBAM attention mechanism according to claim 5, wherein in step (2), the output result of the channel attention mechanism is passed through a spatial attention mechanism, the spatial attention mechanism uses global average pooling, global maximum pooling and global standard deviation pooling to obtain three paths, the three paths are passed through convolution operation after being connected by channels, and finally the final output result is obtained by multiplying the main elements.
7. An intelligent maintenance method for an elevator based on DGConv and improved CBAM attention mechanisms according to claim 1, wherein the intelligent maintenance strategy is generated periodically every week, including the types of faults that may occur in the future week, while prioritizing the types of faults that may occur based on previous historical data.
8. An intelligent maintenance device for an elevator based on a cavity global volume and an improved CBAM attention mechanism, comprising a memory and one or more processors, wherein executable codes are stored in the memory, and wherein the processor implements an intelligent maintenance method for an elevator based on DGConv and an improved CBAM attention mechanism according to any one of claims 1-7 when executing the executable codes.
9. A computer readable storage medium, on which a program is stored, characterized in that the program, when being executed by a processor, implements an intelligent maintenance method of an elevator based on DGConv and improving CBAM attention mechanism as defined in any one of claims 1-7.
10. A computer program product comprising a computer program which, when executed by a processor, implements an intelligent maintenance method of an elevator based on DGConv and improving CBAM attention mechanisms as claimed in any one of claims 1-7.

Description

Elevator intelligent maintenance method and device based on DGConv and CBAM attention mechanism improvement Technical Field The invention relates to the technical field of intelligent maintenance of elevators, in particular to an intelligent maintenance method and device of an elevator based on DGConv and an improved CBAM attention mechanism. Background With the continuous improvement of the automation level in China, the demands on elevators in various areas are increasingly increased, as the elevators gradually enter daily lives of thousands of households, the maintenance work of the elevators is increasingly heavy, the traditional maintenance mode adopts a post maintenance strategy, namely, maintenance personnel are informed to go to the site for investigation and maintenance after the elevators stop due to faults, but in actual use, a plurality of elevator faults belong to accidental faults, namely, the faults trigger a safety protection system of the elevators, but do not cause the loss of the core operation function of the elevators, the elevators can not stop operation normally, or can be automatically recovered quickly even if stopped briefly, such as door lock closing faults, brake band-type brake opening and closing faults and intermittent faults, and for the faults, the elevator maintenance unit usually chooses to ignore only the stop faults. Therefore, the maintenance strategy of the elevator at present only aims at serious elevator stopping faults to carry out key maintenance, the maintenance mode has serious potential safety hazards, the accidental faults do not cause serious elevator stopping, but do not mean no safety risk, the accidental faults cannot simply judge the maintenance priority according to whether the elevator stopping is caused or not, and the factors such as the occurrence frequency of the faults, the duration of the faults and the like are comprehensively considered. Thus, there is an urgent need for an intelligent maintenance strategy for occasional elevator failures. Along with the wide application of the big data technology in industry, the big data technology also shows strong strength in the intelligent maintenance field of the elevator, and is a comprehensive information technology system for collecting, storing, managing and analyzing the data with huge numbers, scattered sources and various formats. The method is characterized in that valuable information and rules are rapidly mined from massive data with low value density by a distributed processing method so as to support decision making, flow optimization and new knowledge discovery. At present, the main board of the elevator has various data types, is not clear in classification, contains discrete or continuous data of various physical quantities, has different data lengths, and is difficult to judge the running state of the elevator only through expert experience. Therefore, the characteristic extraction and information mining of the elevator main board data are realized by utilizing a big data technology, the maintenance plan is dynamically optimized by utilizing the data such as the operation frequency, the history fault, the use environment and the like of the elevator, and the high-risk elevator is early-warned, so that the method is the leading direction in the intelligent maintenance field of the elevator at present. However, a single intelligent maintenance strategy is still a "post maintenance", the principle is that a model is constructed by using historical data for intelligent maintenance of the current fault, and the current requirement of intelligent maintenance of the elevator is to predict the running state of the elevator at the next moment and pertinently propose the intelligent maintenance strategy, so that a method for predicting the state of the elevator at the next moment by using the historical data of the elevator to generate the pertinence intelligent maintenance strategy is urgently designed at present. Disclosure of Invention In order to overcome the defects and the existing problems in the prior art, the invention provides an intelligent maintenance method and device for an elevator based on DGConv and an improved CBAM attention mechanism, wherein DGConv is a hollow global convolution structure, is a brand new convolution structure provided by the invention, CBAM is a convolution attention module which consists of channel attention and space attention, and the improved CBAM attention mechanism provided by the invention adds a standard deviation pooling path on the basis of the traditional CBAM attention mechanism. According to the method, the state of the elevator at the next moment is predicted through historical data in the operation process of the elevator, a targeted intelligent maintenance strategy at the next moment is generated, for example, the probability of door lock faults occurring at the next moment is high when the current moment is predicted, an intelligent maintenance mo