Search

CN-122020378-A - Villa elevator fault monitoring method

CN122020378ACN 122020378 ACN122020378 ACN 122020378ACN-122020378-A

Abstract

The invention discloses a villa elevator fault monitoring method, and aims to solve the defect that the traditional fault monitoring method is difficult to intervene in early fault. The method comprises the steps of constructing a learning model comprising an attention mechanism fusion module and a plurality of base models, collecting multi-source data such as vibration signals, atmospheric pressure, sound signals and the like of each state of an elevator through a sensor to form a data set, dividing the data set into a training set and a test set after preprocessing, inputting the training set into the learning model, independently training the base models, collecting base model results to construct a new feature matrix and an elevator state label, training the new feature matrix, denoising and reconstructing a target elevator signal by using a test set evaluation model, inputting the model to obtain a diagnosis result, uploading the diagnosis result to a database, and notifying operation and maintenance personnel. The method can improve the fault diagnosis accuracy of the villa elevator, reduce false alarm, is suitable for the characteristics of low use frequency and distributed dispersion of the villa elevator, and helps operation and maintenance personnel to grasp operation and maintenance time.

Inventors

  • LI WEIMING
  • Cui Zehui
  • WU JUNLIN
  • JIANG XIANLIANG

Assignees

  • 宁波力隆机电股份有限公司

Dates

Publication Date
20260512
Application Date
20260128

Claims (10)

  1. 1. The villa elevator fault monitoring method is characterized by comprising the following steps: s1, constructing a learning model, wherein the learning model is provided with an attention mechanism fusion module and a plurality of base models; S2, generating a data set, and acquiring data of an elevator car in different running states through a sensor to form the data set, wherein the data information comprises vibration signals, atmospheric pressure and sound signals of the elevator; s3, preprocessing data, and dividing a data set into a training set and a testing set; S4, inputting the training set into a pre-constructed learning model, and independently training each basic model to generate respective training results, wherein the attention mechanism fusion module collects the training results of each basic model and elevator state labels corresponding to samples of the training set as inputs, constructs a new feature matrix and trains to generate final classification results; S5, evaluating the learning model, and testing the learning model by using a test set to enable the accuracy of the learning model on the elevator diagnosis task to reach a preset index; s6, acquiring signals of a target elevator, constructing a multi-domain feature vector set after denoising and reconstruction, and inputting the multi-domain feature vector set into a learning model to obtain a final diagnosis result; and S7, uploading the diagnosis result obtained in the step S6 to a database, comparing, and notifying operation and maintenance personnel according to the requirement.
  2. 2. A villa elevator failure monitoring method according to claim 1, wherein said base model comprises at least one decision tree based model, at least one integrated learning model, at least one support vector machine model and at least one deep learning model.
  3. 3. A villa elevator fault monitoring method as claimed in claim 2, wherein the base models are random forest, adaBoost, SVM and CNN respectively.
  4. 4. A villa elevator fault monitoring method as claimed in claim 3, wherein each basic model is trained by 5-fold cross validation, each basic model outputs 5-dimensional vectors related to the conditions of normal state, loose guide shoes, sudden stop of elevator, unsmooth sliding of car door and abrasion of traction rope, and the output vector combination of each basic model is 4 5 As input to train the metamodel.
  5. 5. The method for monitoring the failure of the villa elevator according to claim 1, wherein the multi-domain feature vector set at least comprises the atmospheric pressure and the time domain, the frequency domain and the entropy features from the denoising reconstruction of the signals.
  6. 6. The villa elevator fault monitoring method according to claim 1, wherein the attention mechanism fusion module comprises a feature extraction layer, an attention marking layer and a weight normalization layer, and the training results of the base models are used as input feature matrix iteration attention network parameters to dynamically adjust the assigned weights of the base models.
  7. 7. A method for monitoring a malfunction of a villa elevator according to claim 1, wherein the denoising reconstruction comprises: performing VMD decomposition, initializing each modal function and the center frequency thereof, iteratively updating the modal components, the center frequency and the Lagrange multiplier through an Alternating Direction Multiplier Method (ADMM) until convergence conditions are met, wherein each modal component is updated in a frequency domain in the iteration process, and the center frequency is recalculated according to energy distribution, wherein the Lagrange multiplier is used for guaranteeing the accuracy of signal reconstruction; And (3) performing false mode rejection, calculating correlation coefficients of each mode and the original signal, and retaining effective components of which the correlation technique is larger than 0.1.
  8. 8. The villa elevator fault monitoring method as claimed in claim 7, wherein the parameters for executing VMD decomposition are obtained through a hawk algorithm, the population scale of the hawk algorithm is set to be 20-50 individuals, the maximum iteration number is 20-100, the value range of K is defined to be an integer of 2-10, the value range of alpha is a continuous value of 1000-5000, the hawk optimization algorithm carries out parameter optimization by simulating four strategies for hawk hunting in nature, the search range is enlarged by adopting a global search strategy in a high-altitude exploration stage, local fine search is realized by a Law flight mechanism in a low-altitude exploration stage, the current optimal solution and population mean are combined for development in a ground attack stage, final optimization is carried out by utilizing a dynamic weight factor in a precise capture stage, and the optimal modal number K and punishment factor alpha parameter combination is output.
  9. 9. The method for monitoring the failure of a villa elevator according to claim 8, wherein the denoising reconstruction further comprises: performing filtering processing on the elevator vibration signal and the elevator sound signal in a frequency range from 70Hz to 75Hz by using a Butterworth band-pass filter, wherein the order of the Butterworth band-pass filter is 3; A Savitzky-Golay filter is used for carrying out moving average processing, the window size is set to be 3, and a first order polynomial is used for fitting data so as to smooth signals.
  10. 10. A method of monitoring a villa elevator malfunction as claimed in claim 1, wherein the data is collected by a sensor having a sampling frequency of 200 Hz.

Description

Villa elevator fault monitoring method Technical Field The invention relates to the field of elevator fault prediction, in particular to a villa elevator fault monitoring method. Background With the acceleration of the urban process and the improvement of the living standard of residents, the villa elevator is used as an important supporting facility of a high-end residence, gradually becomes the first choice of a plurality of families, not only improves the comfort and convenience of living, but also brings great improvement to the daily life of residents. However, safety and reliability of villa elevators are also increasingly concerned. Due to the characteristics of low use frequency, distributed dispersion, long maintenance period and the like, the traditional fault monitoring method is difficult to effectively cope with. Once the elevator fails, people can be trapped, equipment is damaged and even serious safety accidents are caused, and life and property safety of residents is threatened. For example, a failure of the braking system, control system or door system of an elevator may cause a dangerous situation. Patent CN 114933221B-an elevator trouble early warning system discloses an elevator trouble early warning system, including the sensor module, sensor module's output electric connection has communication module, communication module's output signal is connected with the treater, the treater two-way connection has the contrast module, the contrast module two-way connection has the database, treater electric connection has the server, the treater is connected with the maintenance terminal through the server communication, the output electric connection of treater has the siren. According to the elevator control system, the sensor module is arranged to monitor the load weight, the air humidity of the elevator shaft, the distance between the elevator and the elevator door, the vibration of the elevator car, the running sound of the elevator, the air temperature of the elevator shaft, the running speed of the elevator car and the current and voltage of the traction machine in real time, and the acquired information data are compared with the data in the database in real time, so that whether various numerical values are normal in the elevator running process can be accurately judged. However, the mode of directly comparing the vibration signal with the database can only identify the characteristics which obviously meet the fault characteristics, and for distributed household elevators, users can be always in dangerous use stages in the process of repairing the distributed household elevators by feeding back the distributed household elevators to the elevators, so that an efficient and accurate fault monitoring method is needed to ensure the operation safety of villa elevators, so that faults are found earlier when the elevators are in faults, the fault is intervened earlier, and the use time of a dangerous elevator of the users is reduced. Disclosure of Invention The invention overcomes the defect that the prior vibration data obtained by the sensor is directly compared with the database, so that the intervention is difficult in the early stage of the fault, and provides the villa elevator fault monitoring method, which can discover the fault in time, improve the accuracy of the fault diagnosis of the villa elevator, reduce false alarm and help operation and maintenance personnel to grasp the operation and maintenance time. In order to solve the technical problems, the invention adopts the following technical scheme: The villa elevator fault monitoring method specifically comprises the following steps: s1, constructing a learning model, wherein the learning model is provided with an attention mechanism fusion module and a plurality of base models; S2, generating a data set, and acquiring data of an elevator car in different running states through a sensor to form the data set, wherein the data information comprises vibration signals, atmospheric pressure and sound signals of the elevator; s3, preprocessing data, and dividing a data set into a training set and a testing set; S4, inputting the training set into a pre-constructed learning model, and independently training each basic model to generate respective training results, wherein the attention mechanism fusion module collects the training results of each basic model and elevator state labels corresponding to samples of the training set as inputs, constructs a new feature matrix and trains to generate final classification results; S5, evaluating the learning model, and testing the learning model by using a test set to enable the accuracy of the learning model on the elevator diagnosis task to reach a preset index; s6, acquiring signals of a target elevator, constructing a multi-domain feature vector set after denoising and reconstruction, and inputting the multi-domain feature vector set into a learning model to obtain a final diagnosis result; an