Elevator life prediction method based on LSTM-ED
Through the elevator life prediction method based on LSTM-ED, the elevator safety indicators and historical data are used to predict the problem that elevator maintenance in the existing technology is difficult to predict the remaining service life, and higher detection accuracy and accuracy of elevator safety management are achieved.
Patent Information
- Application Number
- CN202411866861.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-18
- Publication Date
- 2025-03-28
AI Technical Summary
The existing technology has the disadvantages of a regular maintenance system in elevator maintenance, and it is difficult to effectively predict the remaining service life of the elevator, resulting in frequent elevator safety accidents.
The elevator life prediction method based on LSTM-ED is adopted. By obtaining the elevator safety indicators, historical data and actual measurement data, time series data is constructed, data preprocessing and training is carried out, and the LSTM-ED network and fully connected network are used for prediction to determine the remaining life and safety level of the elevator.
This method can minimize characteristic errors, improve elevator detection accuracy, effectively predict the remaining service life of the elevator, and improve the accuracy of elevator safety management.
Smart Images
Figure CN119706552A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of elevator life prediction, and in particular to an elevator life prediction method based on LSTM-ED. Background Art
[0002] With the continuous improvement of my country's urbanization level and people's living standards, the number of elevators has increased dramatically and their use has become more frequent. As a means of transportation, elevators have become inseparable from the daily lives of the people, so elevator safety has also become a livelihood project that the people are concerned about and the media pays attention to. In recent years, elevator safety accidents have occurred from time to time, such as elevator eating people, sudden fall of car-type elevators, etc., which have left a great psychological shadow on the people. The frequent occurrence of elevator safety incidents shows that there are still certain drawbacks in my country's existing elevator maintenance system. Elevator maintenance is divided into post-fault maintenance and preventive maintenance. The former is an emergency repair after the elevator fails, and the latter is a preventive maintenance of the elevator by the elevator maintenance department before the failure occurs according to my country's elevator maintenance system, in order to improve the performance of the elevator and extend the service life of the elevator. At present, my country implements a regular maintenance system for elevators. Although the regular preventive maintenance of elevators can avoid the occurrence of elevator safety accidents to a certain extent and extend the remaining service life of elevator safety, it still has certain drawbacks.
[0003] Therefore, it is an urgent problem for technical personnel in this field to propose an elevator life prediction method based on LSTM-ED to solve the problems existing in the prior art. Summary of the invention
[0004] In view of this, the present invention provides an elevator life prediction method based on LSTM-ED, which can minimize feature error and improve elevator detection accuracy.
[0005] In order to achieve the above object, the present invention adopts the following technical solution:
[0006] An elevator life prediction method based on LSTM-ED includes the following steps:
[0007] Data acquisition steps: obtain the elevator's safety indicators, historical data, and actual measurement data, and generate elevator historical data sets and elevator actual measurement data sets;
[0008] Data splitting step: Split the elevator historical data set and the elevator measured data set according to safety indicators to construct historical time series data and measured time series data;
[0009] Data preprocessing step: perform noise reduction on the historical time series data and divide it into multiple subsequences according to the preset window. After preprocessing, the historical time series data forms a training set and a test set.
[0010] Training and verification steps: input the training set into the LSTM-ED network for training to obtain a trained LSTM-ED network model, and input the test set into the trained LSTM-ED network model for verification to obtain a verified model;
[0011] Prediction and classification steps: Input the measured time series data into the verified model, input the obtained data into the fully connected network to predict the remaining life, and determine the elevator safety level based on the remaining life.
[0012] In the above method, optionally, in the data splitting step, the elevator safety index is used as the time series index.
[0013] In the above method, optionally, the data splitting step includes: splitting the elevator historical data set and the elevator measured data set into time series data with weeks as the time unit, and filling or removing the time series with missing data.
[0014] In the above method, optionally, data preprocessing includes: first removing the deviation value of the historical time series data, and performing denoising and normalization processing after using a nonlinear filter function.
[0015] In the above method, optionally, the LSTM-ED network in the data training step includes an encoder, a decoder and a loss calculation unit constructed by LSTM.
[0016] In the above method, optionally, the LSTM encoder converts the input time series into intermediate data for encoding;
[0017] The LSTM decoder decodes the intermediate data in combination with the corresponding time series;
[0018] The loss calculation unit calculates the error between the decoded data and the actual data.
[0019] It can be seen from the above technical solution that compared with the prior art, the present invention provides an elevator life prediction method based on LSTM-ED, which has the following beneficial effects: the present invention uses LSTM-ED to map the original data into the feature space, and finally uses a fully connected network for regression and completion of the prediction, and can minimize the feature error and improve the elevator detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.
[0021] Figure 1This is a flow chart of an elevator life prediction method based on LSTM-ED disclosed in the present invention;
[0022] Figure 2 This is a schematic diagram of an elevator life prediction principle based on LSTM-ED disclosed in the present invention. DETAILED DESCRIPTION
[0023] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0024] In this application, relational terms such as first and second, etc. are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of more restrictions, the elements defined by the sentence "comprise one..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.
[0025] Reference Figure 1 and Figure 2 As shown, the present invention discloses an elevator life prediction method based on LSTM-ED, comprising the following steps:
[0026] Data acquisition steps: obtain the elevator's safety indicators, historical data, and actual measurement data, and generate elevator historical data sets and elevator actual measurement data sets;
[0027] Data splitting step: Split the elevator historical data set and the elevator measured data set according to safety indicators to construct historical time series data and measured time series data;
[0028] Data preprocessing step: perform noise reduction on the historical time series data and divide it into multiple subsequences according to the preset window. After preprocessing, the historical time series data forms a training set and a test set.
[0029] Training and verification steps: input the training set into the LSTM-ED network for training to obtain a trained LSTM-ED network model, and input the test set into the trained LSTM-ED network model for verification to obtain a verified model;
[0030] Prediction and classification steps: Input the measured time series data into the verified model, input the obtained data into the fully connected network to predict the remaining life, and determine the elevator safety level based on the remaining life.
[0031] Furthermore, in the data acquisition step, the elevator data acquired include brake clearance, encoder reading, friction noise, brake acceleration and traction rope wear, and the safety indicators are the brake clearance, encoder reading, friction noise, braking force and traction rope factory data of the elevator in the initial state.
[0032] Specifically, during the operation of the elevator, when the elevator car arrives at the station where it needs to stop, the traction machine decelerates to an extremely low speed according to the set program, and the electromagnetic coil is de-energized. Under the pressure of the spring, the brake shoe holds the brake wheel again, and uses the friction of the friction plate to stop the traction machine. It can be seen that the main reasons for the brake decay during the operation of the elevator are too small braking torque, too large brake air gap and braking distance, and wear of the friction plate. At the same time, since the traction rope is connected to the car, after a period of use, it usually has internal and external wear, broken wires, fatigue, rust, etc., which are all regarded as the wear degree of the traction rope.
[0033] Furthermore, in the data splitting step, the elevator safety index is used as the time series index.
[0034] Furthermore, the data splitting step includes: splitting the elevator historical data set and the elevator measured data set into time series data with weeks as the time unit, and filling or removing the time series with missing data.
[0035] Specifically, when a time series is missing a large amount of data, the data or time series is deleted accordingly. When a time series is missing some data, rolling statistics or interpolation methods are used to fill in the missing data. Rolling statistics include: simple moving average, weighted moving average and exponential moving average; interpolation includes: linear interpolation, spline interpolation and time interpolation.
[0036] Furthermore, data preprocessing includes: first removing the deviation values of the historical time series data, and using a nonlinear filter function to perform denoising and then normalizing.
[0037] Specifically, the filtering method includes using a bandpass filter to perform corresponding window filtering, including a Hamming window, a Blackman window and a Kaiser window. Let L be the data length, the data length after preprocessing is l, and the number of data obtained is L-l+1.
[0038] Furthermore, the LSTM-ED network in the data training step includes an encoder, a decoder, and a loss calculation unit constructed by LSTM.
[0039] Furthermore, the LSTM encoder converts the input time series into intermediate data for encoding;
[0040] The LSTM decoder decodes the intermediate data in combination with the corresponding time series;
[0041] The loss calculation unit calculates the error between the decoded data and the actual data.
[0042] Specifically, given the initial time series X, the input data is obtained after preprocessing Input it into the LSTM encoder and convert it into intermediate data During the decoding process, the decoder reconstructs the time series in the opposite order, and inputs the parameters at time t to time t-1. The decoder combines the parameters at the previous time and the current intermediate data to reconstruct a time series with the same length as the initial time series. The loss calculation unit adopts Among them, t h This is the working state of the elevator at the initial stage of operation. To output data for the decoding unit, the loss function is optimized and updated to obtain the reconstructed feature value to reflect the performance attenuation of the device. After obtaining the optimal weight, the weight of the LSTM-ED neural network is fixed unchanged, the time series is input, and the reconstructed feature output is obtained based on the trained neural network.
[0043] Furthermore, the reconstructed eigenvalues are input into the DNN network, the eigenvector at each moment is used as the neural network input, and the loss function is used for regression to determine the life of the elevator.
[0044] Furthermore, elevator safety is divided into five levels. The national standard elevator life is generally 15 years. The remaining 10-15 years in the predicted life level are divided into five levels, 7-10 years are divided into four levels, 4-7 years are divided into three levels, 2-4 years are divided into two levels, and 1-2 years are divided into one level.
[0045] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can refer to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment. The system and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without creative work.
[0046] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. An elevator life prediction method based on LSTM-ED, characterized in that: The following steps are included Data acquisition steps: obtain the elevator's safety indicators, historical data, and actual measurement data, and generate elevator historical data sets and elevator actual measurement data sets; Data splitting step: Split the elevator historical data set and the elevator measured data set according to safety indicators to construct historical time series data and measured time series data; Data preprocessing step: perform noise reduction on the historical time series data and divide it into multiple subsequences according to the preset window. After preprocessing, the historical time series data forms a training set and a test set. Training and verification steps: input the training set into the LSTM-ED network for training to obtain a trained LSTM-ED network model, and input the test set into the trained LSTM-ED network model for verification to obtain a verified model; Prediction and classification steps: Input the measured time series data into the verified model, input the obtained data into the fully connected network to predict the remaining life, and determine the elevator safety level based on the remaining life.
2. The elevator life prediction method based on LSTM-ED according to claim 1 is characterized in that: In the data splitting step, elevator safety indicators are used as time series indicators.
3. The elevator life prediction method based on LSTM-ED according to claim 1 is characterized in that: The data splitting step includes: splitting the elevator historical data set and the elevator measured data set into time series data with weeks as the time unit, and filling or removing the time series with missing data.
4. The elevator life prediction method based on LSTM-ED according to claim 1 is characterized in that: Data preprocessing includes: first removing the deviation values of historical time series data, and then using nonlinear filtering function to denoise and then normalize.
5. The elevator life prediction method based on LSTM-ED according to claim 1 is characterized in that: The LSTM-ED network in the data training step includes an encoder, a decoder, and a loss calculation unit constructed by LSTM.
6. The elevator life prediction method based on LSTM-ED according to claim 5 is characterized in that: The LSTM encoder converts the input time series into intermediate data for encoding; The LSTM decoder decodes the intermediate data in combination with the corresponding time series; The loss calculation unit calculates the error between the decoded data and the actual data.