Elevator running state monitoring and early warning method and platform based on multi-source data fusion

Abstract

The invention relates to the technical field of data processing, and discloses an elevator running state monitoring and early warning method and platform based on multi-source data fusion. The method comprises the steps that elevator operation data are collected through a multi-source sensor, an improved entropy weight method is introduced to fuse the data, a mixed deep learning model is constructed to predict the elevator state, parameter abnormity is analyzed based on association rules, graded early warning and fault diagnosis reports are generated, and intelligent monitoring and early warning of the elevator are achieved. Accurate collection, deep analysis and fault prediction of elevator operation data are achieved, and therefore scientificity and accuracy of elevator maintenance are improved.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and in particular, to an elevator operation status monitoring and early warning method and platform based on multi-source data fusion. Background Art

[0002] With the acceleration of the urbanization process and the continuous increase in building height, elevators, as indispensable vertical transportation tools in modern buildings, their safe operation is crucial for people's daily life and work. Traditional elevator operation status monitoring mainly relies on regular manual inspections and simple parameter monitoring. Maintenance personnel evaluate the elevator operation status and potential failure risks through experience and inspections at fixed intervals. Most existing elevator monitoring systems use single-sensor data collection and simple threshold alarm methods, making it difficult to comprehensively capture the complex operation characteristics and potential failure signals of elevators.

[0003] There are many limitations in existing elevator monitoring technologies: the data collection method is single, and the deep fusion of multi-source data cannot be achieved; fault diagnosis relies on experience and lacks data-driven intelligent analysis; the ability of predictive maintenance is weak, and potential faults cannot be accurately located and warned; the recommended maintenance plan lacks systematicness and scientificity, and the maintenance efficiency and accuracy are low. These technical deficiencies lead to high elevator maintenance costs and difficult effective control of failure risks, seriously restricting the improvement of elevator operation safety and service life. Summary of the Invention

[0004] This application provides an elevator operation status monitoring and early warning method and platform based on multi-source data fusion, which is used to achieve accurate collection, in-depth analysis, and fault prediction of elevator operation data, thereby improving the scientificity and accuracy of elevator maintenance.

[0005] In a first aspect, this application provides an elevator operation status monitoring and early warning method based on multi-source data fusion. The elevator operation status monitoring and early warning method based on multi-source data fusion includes: collecting and preprocessing vibration signals, noise, speed fluctuations, motor temperature, load data, door machine parameters, and guide rail deformation data during elevator operation to obtain normalized multi-source data; performing fusion processing on the normalized multi-source data based on an improved entropy weight method, adjusting the weight distribution by introducing a sensitivity correction factor to obtain an elevator operation status fusion index; inputting the elevator operation status fusion index and the original multi-source data into a hybrid deep learning model composed of a bidirectional long short-term memory network, an attention mechanism, and a convolutional neural network for training to obtain an elevator status prediction model; performing a correlation analysis on the parameter status and fault types according to the output result of the elevator status prediction model to obtain a set of association rules for parameter anomaly combinations and fault types; evaluating the elevator operation status based on the set of association rules to obtain hierarchical early warning information, and generating a fault diagnosis report based on the hierarchical early warning information.

[0006] In a second aspect, the present application provides an elevator operation status monitoring and early warning platform based on multi-source data fusion. The elevator operation status monitoring and early warning platform based on multi-source data fusion includes:

[0007] A processing module, configured to collect and preprocess vibration signals, noise, speed fluctuations, motor temperature, load data, door machine parameters, and guide rail deformation data during the operation of the elevator to obtain normalized multi-source data;

[0008] A fusion module, configured to perform fusion processing on the normalized multi-source data based on an improved entropy weight method, and adjust the weight distribution by introducing a sensitivity correction factor to obtain an elevator operation status fusion index;

[0009] An input module, configured to input the elevator operation status fusion index and the original multi-source data into a hybrid deep learning model composed of a bidirectional long short-term memory network, an attention mechanism, and a convolutional neural network for training processing to obtain an elevator status prediction model;

[0010] An analysis module, configured to perform a correlation analysis on parameter status and fault types according to the output result of the elevator status prediction model to obtain a set of association rules for parameter anomaly combinations and fault types;

[0011] An evaluation module, configured to evaluate the elevator operation status based on the set of association rules to obtain hierarchical early warning information, and generate a fault diagnosis report based on the hierarchical early warning information.

[0012] In a third aspect, a computer device is provided, including: a memory and at least one processor, wherein instructions are stored in the memory; the at least one processor calls the instructions in the memory to enable the computer device to execute the above-mentioned elevator operation status monitoring and early warning method based on multi-source data fusion.

[0013] In a fourth aspect, a computer-readable storage medium is provided, wherein instructions are stored in the computer-readable storage medium, and when the instructions are run on a computer, the computer is enabled to execute the above-mentioned elevator operation status monitoring and early warning method based on multi-source data fusion.

[0014] In the technical solution provided by this application, by innovatively integrating technical features such as multi-source sensor data acquisition, improved entropy weight data fusion, hybrid deep learning model construction, and intelligent fault correlation analysis, the scientific nature and accuracy of elevator operation status monitoring have been significantly improved. At the data acquisition level, through the collaborative work of various sensors such as acceleration sensors, sound sensors, and speed sensors, comprehensive capture of multi-dimensional features such as vibration signals, noise, speed fluctuations, motor temperature, and load data during elevator operation has been realized, laying a solid data foundation for subsequent in-depth analysis. The improved entropy weight method introduces a sensitivity correction factor, breaking through the limitations of weight allocation in the traditional entropy weight method, and being able to adjust the weights of different data sources more dynamically and accurately, effectively balancing the importance of various sensor data. The innovation of the hybrid deep learning model lies in integrating bidirectional long short-term memory networks, attention mechanisms, and convolutional neural networks, giving full play to the advantages of artificial intelligence algorithms in processing complex time-series data. Specifically, the bidirectional LSTM network captures long-term dependencies, the attention mechanism realizes dynamic weight allocation for key time nodes, and the one-dimensional convolutional neural network extracts local feature patterns. The organic combination of the three algorithms significantly improves the accuracy and robustness of elevator operation status prediction. The introduction of association rule mining technology realizes in-depth correlation analysis between parameter abnormal states and fault types, and constructs an intelligent early warning system for elevator faults by setting dynamic thresholds and time-series information weighting. The fault case retrieval and maintenance plan recommendation links adopt feature vector similarity matching and multi-dimensional scoring mechanisms, breaking through the traditional experience-driven maintenance mode and establishing a data-driven scientific maintenance paradigm. This technical path of multi-source data fusion and artificial intelligence algorithm empowerment significantly improves the accuracy and early warning ability of elevator operation status monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for description in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0016] Figure 1 It is a schematic diagram of an embodiment of the elevator operation status monitoring and early warning method based on multi-source data fusion in the embodiments of this application;

[0017] Figure 2 It is a schematic diagram of an embodiment of the elevator operation status monitoring and early warning platform based on multi-source data fusion in the embodiments of this application;

[0018] Figure 3 It is a schematic block diagram of the structure of a computer device in the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] The embodiments of the present application provide an elevator operation status monitoring and early warning method and platform based on multi-source data fusion. Terms such as "first", "second", "third", "fourth", etc. (if any) in the description, claims and the above-mentioned drawings of the present application are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily limit to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0020] For ease of understanding, the specific process of the embodiments of the present application is described below. Please refer to Figure 1 , an embodiment of the elevator operation status monitoring and early warning method based on multi-source data fusion in the embodiments of the present application includes:

[0021] Step S101: Collect and preprocess vibration signals, noise, speed fluctuations, motor temperature, load data, door machine parameters, and guide rail deformation data during the operation of the elevator to obtain normalized multi-source data;

[0022] Step S102: Perform fusion processing on the normalized multi-source data based on the improved entropy weight method, and adjust the weight distribution by introducing a sensitivity correction factor to obtain an elevator operation status fusion index;

[0023] Step S103: Input the elevator operation status fusion index and the original multi-source data into a hybrid deep learning model composed of a bidirectional long short-term memory network, an attention mechanism, and a convolutional neural network for training to obtain an elevator status prediction model;

[0024] Step S104: According to the output result of the elevator status prediction model, perform a correlation analysis on the parameter status and the fault type to obtain a set of association rules between parameter anomaly combinations and fault types;

[0025] Step S105: Evaluate the elevator operation status based on the set of association rules to obtain hierarchical early warning information, and generate a fault diagnosis report based on the hierarchical early warning information.

[0026] It can be understood that the execution subject of the present application can be an elevator operation status monitoring and early warning platform based on multi-source data fusion, or a terminal or a server. Specifically, it is not limited here. The embodiments of the present application are described by taking the server as the execution subject as an example.

[0027] Specifically, elevator operation status data is obtained through multiple types of sensors. Acceleration sensors are installed at the bottom, top, and around the elevator car, with a sampling frequency set to 200 Hz to record the vibration acceleration values of the XYZ axes; the temperature sensor records the surface temperature of the motor every 10 seconds; the speed sensor records the change in the up and down speed of the elevator at a sampling frequency of 50 Hz; the door machine control system records the time and current change of each door opening and closing cycle; the load sensor monitors the change in the weight inside the car in real time; the sound sensor records the noise decibel value inside the car at a sampling frequency of 20 Hz; the displacement sensor measures the change in the gap between the guide rail and the car. The collected raw data is preprocessed. First, the wavelet transform method is used to eliminate environmental interference, and then the missing data is filled in by linear interpolation or forward filling method. Then, different types of data are uniformly normalized to the interval [-1, 1] to eliminate the influence of dimensions, obtaining normalized multi-source data. The normalized multi-source data is fused. A standardized data matrix is constructed, processed according to the nature of positive and negative indicators respectively, the proportion of each indicator at each sampling point is calculated, and then the information entropy is calculated. The difference coefficient is calculated through the information entropy. The larger the difference coefficient, the greater the role of the indicator. On the basis of the traditional entropy weight method, a sensitivity correction factor is introduced to adjust the weight distribution. The sensitivity correction factor is calculated by adding 1 to the ratio of the standard deviation of the indicator to the maximum standard deviation. After the final indicator weight is corrected by the correction factor, the elevator operation status fusion index at each sampling point is calculated, with a value range of [0, 1]. The closer the value is to 1, the better the elevator operation status. Then, the elevator operation status fusion index and the original multi-source data are input into the hybrid deep learning model for training. The input window length is set to 24 hours, and the prediction window length is set to 8 hours. The hybrid deep learning model includes a bidirectional long short-term memory network layer, an attention mechanism layer, a one-dimensional convolutional neural network layer, and a fully connected layer. The bidirectional long short-term memory network layer captures the long-term dependence relationship of time series data; the attention mechanism layer identifies important time points and features; the one-dimensional convolutional neural network layer extracts local feature patterns; the fully connected layer fuses the extracted features. The mean squared error is used as the loss function during the training process, and the Adam optimizer is used for parameter optimization. When the loss of the validation set does not decrease for 10 consecutive cycles, the training is stopped in advance to prevent overfitting. Based on the trained elevator state prediction model, the correlation analysis between parameter status and fault types is carried out. The predicted parameter values are discretized into three levels: normal state, slightly abnormal state, and severely abnormal state. A transaction database is established in combination with historical fault records, and the improved Apriori algorithm is used to discover frequent item sets, forming the association rules between parameter anomalies and fault types. The strength of the rules is evaluated by calculating three indicators: support, confidence, and lift, and a time series information weighting mechanism is introduced to assign different weights to the data at different time points, forming a set of association rules between parameter anomalies and fault types. The elevator operation status is evaluated based on the set of association rules. An adaptive dynamic threshold system is constructed, and the threshold calculation considers factors such as elevator type, service life, and load status.According to the matching results of association rules, the elevator operation status is divided into four levels: normal, attention, warning, and danger, forming hierarchical early warning information. Calculate the probabilities of each potential fault type, analyze the development trend of high-probability faults, and estimate the fault occurrence time. Retrieve similar historical cases from the fault knowledge base and extract maintenance experience to form a reference maintenance plan. Finally, integrate the early warning level, fault probability, time estimation results, and maintenance plan into a fault diagnosis report.

[0028] Taking the elevator in a certain shopping mall as an example, abnormal fluctuations in the XYZ-axis vibration acceleration values are collected through an acceleration sensor, and the vibration frequency is 40% higher than the normal value; the temperature sensor shows that the motor temperature continues to rise to 65°C; the noise sensor records that the noise during the car operation reaches 75 decibels; the displacement sensor measures that the guide rail clearance becomes larger. After preprocessing and fusing these data, the elevator operation status fusion index drops to 0.37, which is at the "poor" level. The prediction model predicts that the fusion index will further drop to 0.22 within 24 hours. Association rule analysis shows that the current parameter combination has a high matching degree with the "guide rail wear" fault type, with a support of 0.045, a confidence of 0.86, and a lift of 12.3. The early warning platform generates a "warning" level early warning and recommends checking the guide rail status and arranging maintenance first in the fault diagnosis report.

[0029] In the embodiments of the present application, by innovatively integrating technical features such as multi-source sensor data acquisition, improved entropy weight data fusion, hybrid deep learning model construction, and intelligent fault correlation analysis, the scientific nature and accuracy of elevator operation status monitoring have been significantly improved. At the data acquisition level, through the collaborative work of multiple sensors such as acceleration sensors, sound sensors, speed sensors, etc., comprehensive capture of multi-dimensional features such as vibration signals, noise, speed fluctuations, motor temperature, load data, etc. during the elevator operation has been achieved, laying a solid data foundation for subsequent in-depth analysis. The improved entropy weight method introduces a sensitivity correction factor, breaking through the limitations of the weight allocation of the traditional entropy weight method, and being able to adjust the weights of different data sources more dynamically and accurately, effectively balancing the importance of various sensor data. The innovation of the hybrid deep learning model lies in integrating the bidirectional long short-term memory network, attention mechanism, and convolutional neural network, giving full play to the advantages of artificial intelligence algorithms in processing complex time-series data. Specifically, the bidirectional LSTM network captures long-term dependence relationships, the attention mechanism realizes dynamic weight allocation for key time nodes, and the one-dimensional convolutional neural network extracts local feature patterns. The organic combination of the three algorithms significantly improves the accuracy and robustness of elevator operation status prediction. The introduction of association rule mining technology realizes in-depth correlation analysis between parameter abnormal states and fault types. By setting dynamic thresholds and time-series information weighting, an intelligent early warning system for elevator faults is constructed. The fault case retrieval and maintenance plan recommendation links adopt feature vector similarity matching and multi-dimensional scoring mechanisms, breaking through the traditional experience-driven maintenance mode and establishing a data-driven scientific maintenance paradigm. This technical path of multi-source data fusion and artificial intelligence algorithm empowerment significantly improves the accuracy and early warning ability of elevator operation status monitoring.

[0030] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0031] (1) Collect original operation data through a variety of sensor devices set in the elevator system, including collecting vibration signals through acceleration sensors, collecting noise data through sound sensors, collecting speed fluctuation values through speed sensors, collecting motor temperature through temperature sensors, collecting load data through load sensors, collecting door machine parameters through the door machine control system, and collecting guide rail deformation data through displacement sensors, to obtain the original monitoring data of elevator operation;

[0032] (2) Eliminate environmental interference from the original monitoring data of elevator operation through the wavelet transform method, and separate the effective signal and noise according to the directional noise reduction algorithm to obtain the noise-reduced data of elevator operation;

[0033] (3) Supplement the missing data based on the elevator operation noise reduction data through linear interpolation or forward filling method, and fill the data blank points according to the principle of time series continuity to obtain the complete elevator operation data with a completeness of not less than 99.5%.

[0034] (4) Synchronize the complete elevator operation data through the timestamp alignment function, and reconstruct the data streams of each sensor according to the unified sampling period to obtain the elevator monitoring data with time series alignment.

[0035] (5) Identify and replace the outlier points in the elevator monitoring data with time series alignment by using the 3σ principle, and screen out the data points deviating from the normal range based on the statistical distribution characteristics to obtain the elevator monitoring data after outlier processing.

[0036] (6) Convert the elevator monitoring data after outlier processing to the interval [-1, 1] according to the maximum-minimum normalization method to eliminate the influence of different physical dimensions and obtain the normalized multi-source data.

[0037] Specifically, a variety of sensor devices are set in the elevator system, including acceleration sensors, sound sensors, speed sensors, temperature sensors, load sensors, door machine control systems, and displacement sensors. The acceleration sensors are installed at the bottom, top, and around the elevator car, with a sampling frequency of 200 Hz, recording the vibration acceleration values of the XYZ axes to form vibration time series data; the sound sensors are installed inside the car, with a sampling frequency of 20 Hz, continuously recording the noise decibel values inside the car; the speed sensors are installed in the elevator drive system, with a sampling frequency of 50 Hz, monitoring the up and down speeds of the elevator and their fluctuations; the temperature sensors are attached to the surface of the motor, recording the temperature data every 10 seconds; the load sensors are installed at the bottom of the car, real-time monitoring the weight changes inside the car; the door machine control system records the time and current changes of each door opening and closing; the displacement sensors measure the changes in the gap between the guide rail and the car. These sensors work simultaneously and continuously collect data to form the original multi-dimensional and multi-angle elevator operation monitoring data. After obtaining the original elevator operation monitoring data, noise reduction processing is required to eliminate environmental interference. The wavelet transform method is used to denoise the data. The wavelet transform has the characteristics of time-frequency localization and is suitable for processing non-stationary signals. The specific steps include selecting a suitable wavelet basis function, decomposing the original signal into wavelet coefficients of different frequencies, setting a threshold to screen the noise coefficients, retaining the effective signal coefficients, and then reconstructing the signal. The directional noise reduction algorithm selects a suitable wavelet basis function and threshold strategy according to the characteristics of different types of sensors. The Daubechies wavelet is selected for the elevator vibration signal, the Symlet wavelet is selected for the noise data, and the Coiflet wavelet is selected for the slowly varying signals such as temperature. The signal is decomposed into multiple frequency sub-bands through wavelet decomposition, the soft threshold or hard threshold method is applied to the high-frequency sub-bands to suppress the noise, and the information of the low-frequency sub-bands is retained. Finally, the elevator operation noise reduction data is reconstructed.

[0038] During the data collection process, data may be missing due to reasons such as temporary sensor failures, communication interruptions, or power fluctuations. To handle the missing values in the elevator operation noise reduction data, linear interpolation or forward filling method is used. Linear interpolation is applicable to the situation where the data changes smoothly before and after the missing point, and the missing value is calculated through the known data points on both sides of the missing point; the forward filling method is applicable to the situation of sudden missing, and the nearest valid value before the missing point is used for filling. For different types of data, appropriate filling methods are selected. Since the vibration signal and noise data change rapidly, linear interpolation is adopted; the temperature and load data change slowly, and forward filling is used. Through the principle of time series continuity, the time correlation of the data is analyzed, and the possible values of the missing data points are reasonably inferred, and finally the complete elevator operation data with a integrity of not less than 99.5% is formed. Due to the different sampling frequencies of each sensor, there are differences in the time points of data collection, and time synchronization processing needs to be performed on the complete elevator operation data. Through the timestamp alignment function, the data of different sensors are reorganized according to a unified time reference. First, the synchronous reference time point is determined, and then the data of each sensor is resampled to conform to a unified sampling period. For the vibration signal and speed data with high sampling rates, downsampling processing is performed; for the data with low sampling rates such as temperature and door machine parameters, interpolation processing is performed. The least common multiple principle is used to determine the unified sampling period, and then all sensor data are mapped to this unified time axis to form the elevator monitoring data with time series alignment.

[0039] Outlier processing is performed on the elevator monitoring data with time series alignment. The 3σ principle in statistics is adopted, that is, under the normal distribution, the probability that the data falls within the range of the mean μ±3σ is 99.73%, and the data points outside this range are regarded as outliers. For each type of sensor data, its mean μ and standard deviation σ are calculated, and the upper and lower limit thresholds μ±3σ are set to detect the data points exceeding the thresholds. For the identified outliers, the median replacement method or the local mean replacement method is used for processing. The median replacement method is applicable to the situation with more outliers, and the local mean replacement is applicable to isolated outliers. Through outlier processing, unreasonable data disturbances are removed, and the data reflecting the true operation state of the elevator is retained to form the elevator monitoring data after outlier processing. Normalization processing is performed on the elevator monitoring data after outlier processing. Since the physical dimensions of the data of different sensors are different and the numerical ranges vary greatly, they need to be unified to the same scale. The maximum-minimum normalization method is adopted to linearly map various types of data to the interval [-1,1]. Through normalization processing, the influence of different physical dimensions is eliminated, so that all data can be compared and fused on the same scale to form normalized multi-source data.

[0040] Taking the passenger elevator in a certain high-rise building as an example, the vibration data collected by the acceleration sensor contains the noise interference of the building's own vibration. The Daubechies4 wavelet is used for 5-layer decomposition, and the soft threshold method is used to process the high-frequency coefficients. After reconstruction, a pure elevator vibration signal is obtained. During the data collection process, due to poor contact, the temperature sensor causes intermittent data loss. For the missing values at 10 consecutive time points, linear interpolation is used to fill them, making the data integrity reach 99.7%. The sampling frequencies of each sensor vary from 20Hz to 200Hz. By setting a unified sampling interval of 100ms, all data is resampled and time-aligned. In the motor temperature data, there is an abnormal value that suddenly rises to 120°C, which significantly exceeds the normal operating temperature range (30 - 80°C). The calculated mean is 45°C and the standard deviation is 5°C. The abnormal values exceeding 60°C are replaced with the average of the previous and subsequent time points. The vibration data in the range of ±20g, the temperature data in the range of 30 - 80°C, and the load data in the range of 0 - 1600kg are all uniformly normalized to the [-1, 1] interval to form a standardized multi-source data set.

[0041] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0042] (1) Construct a standardized data matrix for the normalized multi-source data. Through the normalization calculation of positive and negative indicators, a standardized processing matrix is obtained;

[0043] (2) Calculate the proportion and information entropy of each indicator at each sampling point according to the standardized processing matrix. Through the operation of multiplying the proportion by the logarithm and then summing, the information entropy of each indicator is obtained;

[0044] (3) Calculate the difference coefficient and initial weight based on the information entropy of each indicator. The difference coefficient is obtained by subtracting the information entropy from the constant 1, and the initial weight is obtained by dividing the difference coefficient by the sum, obtaining the initial weight configuration;

[0045] (4) Calculate the sensitivity correction factor through the ratio of the standard deviation of each indicator to the maximum standard deviation, and add 1 to the ratio as the correction factor to obtain the weight adjustment parameter;

[0046] (5) Use the weight adjustment parameter to correct the initial weight. By multiplying the correction factor by the initial weight and normalizing, the final weight is obtained;

[0047] (6) Based on the weighted sum of the final weight and the standardized data, establish a sub-scenario evaluation system for the elevator operation under different working conditions, and obtain the elevator operation state fusion index.

[0048] Specifically, the normalized multi-source data is fused, and a comprehensive evaluation index for the elevator operation status is constructed by improving the entropy weight method. First, a standardized data matrix is constructed for the normalized multi-source data, and the elevator operation parameters are divided into two categories: positive indicators and negative indicators. Positive indicators are parameters where the larger the value, the better the elevator status, such as the smoothness of the door machine operation; negative indicators are parameters where the smaller the value, the better the elevator status, such as vibration amplitude and noise decibel value. For positive indicators, the following formula is used for standardization:

[0049]

[0050] For negative indicators, the following formula is used for standardization:

[0051]

[0052] where x ij represents the value of the j-th indicator at the i-th sampling point, min(x j ) and max(x j ) represent the minimum and maximum values of the j-th indicator respectively, and x' ij represents the value after standardization. Through standardization, all indicator values are mapped to the interval [0,1], forming a standardized processing matrix.

[0053] Then, according to the standardized processing matrix, calculate the proportion of each indicator at each sampling point. The proportion calculation formula is:

[0054]

[0055] where p ij represents the proportion of the j-th indicator at the i-th sampling point, and S represents the total number of sampling points. Then calculate the information entropy of each indicator. Information entropy is an indicator to measure data uncertainty, and the calculation formula is:

[0056]

[0057] where E j represents the information entropy of the j-th indicator, κ = 1 / ln(S) is a normalization constant to ensure that the value of E j is between 0 and 1. When p ij = 0, it is defined that p ij ln(p ij ) = 0. The larger the information entropy, the lower the discrimination degree of the indicator for the elevator status.

[0058] Based on the information entropy of each indicator, calculate the difference coefficient. The difference coefficient represents the discrimination ability of the indicator, and the calculation formula is:

[0059] d j = 1 - E j

[0060] where d j represents the coefficient of variation of the j-th index. The larger the coefficient of variation, the greater the role of the index in the evaluation. Further calculate the initial weight, and the initial weight calculation formula is:

[0061]

[0062] where w j represents the initial weight of the j-th index, and N represents the total number of indexes. After the initial weight configuration is completed, each index obtains the corresponding weight value.

[0063] The traditional entropy weight method only considers the data distribution characteristics and lacks the consideration of the index sensitivity. Therefore, a sensitivity correction factor is introduced for adjustment. The sensitivity correction factor calculation formula is:

[0064]

[0065] where α j represents the sensitivity correction factor of the j-th index, σ j represents the standard deviation of the j-th index, and σ max represents the maximum value of the standard deviations of all indexes. The index with a large standard deviation is more sensitive to data changes and is given a higher weight.

[0066] Use the sensitivity correction factor to correct the initial weight, and the correction formula is:

[0067]

[0068] where w j ' represents the final weight of the j-th index. Through sensitivity correction, the role of the index sensitive to the elevator state change is more prominent, and the sensitivity of the fusion index is improved.

[0069] Finally, based on the final weight and the standardized data, perform weighted summation to calculate the elevator operation state fusion index. The calculation formula is:

[0070]

[0071] where F i represents the elevator operation state fusion index of the i-th sampling point, and the value range is [0, 1]. The value closer to 1 indicates that the elevator operation state is better, and the value closer to 0 indicates a higher potential failure risk.

[0072] For different elevator operation conditions, such as no-load upward, full-load upward, no-load downward, full-load downward, door opening and closing process, etc., separately divide the sub-scenarios and construct the fusion index to form a complete elevator operation state evaluation system.

[0073] Taking an elevator in a certain shopping mall as an example, seven types of index data such as vibration signals, noise, speed fluctuations, motor temperature, load data, door machine parameters, and guide rail deformation were obtained through multi-source data collection. Among them, vibration amplitude, noise decibel value, speed volatility, motor temperature, and guide rail deformation belong to negative indicators, while load stability and door machine smoothness belong to positive indicators. Standardize each indicator. For example, the vibration amplitude is standardized from 0.1g to 0.5g to the 0-1 interval. Calculate the information entropy of each indicator. It is found that the information entropy of the motor temperature is 0.45, the information entropy of the vibration signal is 0.32, and the information entropy of the noise data is 0.38, indicating that the vibration signal has the highest discrimination. Calculate the coefficient of variation. The coefficient of variation of the vibration signal is 0.68, and the coefficient of variation of the motor temperature is 0.55, obtaining the initial weight configuration. The standard deviation of the vibration signal is 0.08, and the standard deviation of the motor temperature is 0.05. Calculate the sensitivity correction factor to correct the initial weight. Finally, through weighted summation, the elevator operation status fusion index is 0.82, indicating that the elevator operation status is good. When the elevator enters the full-load upward working condition, the weight of the load data increases, and the fusion index drops to 0.76, entering the "good" level. When abnormal door machine parameters are detected, the corresponding weight increases, and the fusion index further drops to 0.63, triggering a "caution" level warning.

[0074] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0075] (1) Divide the historical time series data by a sliding window, set an input duration of 24 hours and a prediction duration of 8 hours to form a training sample set;

[0076] (2) Input the elevator operation status fusion index and the original multi-source data in the training sample set into the bidirectional long short-term memory network layer, capture long-term dependence relationships through forward propagation and backward propagation, and obtain sequence feature representations;

[0077] (3) Construct an attention weight coefficient based on the sequence feature representation, calculate the importance distribution of each time point through the softmax function, and obtain a weighted feature representation;

[0078] (4) Perform local feature extraction on the weighted feature representation through a one-dimensional convolutional neural network, and use a sliding convolutional kernel to capture local pattern changes to obtain a feature mapping result;

[0079] (5) Input the feature mapping result into the fully connected layer for feature fusion, and at the same time introduce the dropout mechanism to control the risk of overfitting to obtain a predicted output vector;

[0080] (6) Calculate the mean square error loss according to the predicted output vector and the true label, update the parameters through the Adam optimizer, and stop training when the loss on the validation set no longer decreases to obtain an elevator state prediction model.

[0081] Specifically, the historical time series data is divided into sliding windows. The input window length is set to 24 hours, and the prediction window length is set to 8 hours. The sliding window is a time series processing technique that divides a continuous time series into multiple overlapping data segments by sliding a fixed-size window along the time axis. In specific implementation, for the continuously collected elevator operation data, every 24 hours of data is taken as input features, and the corresponding data for the next 8 hours is taken as the prediction target. The sliding step size can be set to 1 hour, that is, every time it moves forward 1 hour, a new training sample is generated. In this way, a large number of training samples can be generated from long-term historical data. For example, more than 8,000 training samples can be generated from one year's data, forming a training sample set containing input features and prediction targets. The elevator operation state fusion index and the original multi-source data in the training sample set are input into the bidirectional long short-term memory network layer for processing. The bidirectional long short-term memory network (BiLSTM) is a special recurrent neural network that can consider both past and future information and is suitable for processing time series data. BiLSTM consists of a forward LSTM and a backward LSTM. The forward LSTM processes the sequence from left to right, and the backward LSTM processes the sequence from right to left. Through this bidirectional structure, long-term dependencies in the time series data can be captured. In elevator state prediction, the BiLSTM layer receives an input tensor with dimensions [batch size × time step × number of features]. Each time step corresponds to one hour of data, including the fusion index and the original multi-source data. Information is screened and updated through a gating mechanism (including an input gate, a forget gate, and an output gate). The input gate controls the degree to which new information enters the cell state, the forget gate controls the degree to which old information is retained, and the output gate controls the degree to which the cell state is passed to the output. Two layers are set for the BiLSTM layer, and the number of hidden units is 128. Parameters are learned through forward propagation and backward propagation algorithms, and finally sequence feature representations are output to capture the time series patterns and long-term trends of the elevator operation data.

[0082] Construct the attention weight coefficients based on the sequence feature representation output by the BiLSTM. The attention mechanism is a technique that allows the model to selectively focus on the important parts of the input sequence. In elevator state prediction, the data at different time points contribute differently to the prediction result. Through the attention mechanism, the weights of the data at each time point can be dynamically adjusted. When specifically implemented, first, the hidden state output by the BiLSTM is linearly transformed, then the tanh activation function is used for non-linear mapping, and then the softmax function is used for normalization to obtain the attention weights at each time point. The softmax function converts the input into a probability distribution, ensuring that the sum of all weights is 1. The attention weights reflect the importance of the data at each time point to the prediction. The larger the value, the greater the impact of that time point on the prediction result. Through the weighted sum of the attention weights and the original features, the weighted feature representation is obtained, highlighting the information of important time points and suppressing the noise interference of irrelevant time points.

[0083] Perform local feature extraction on the weighted feature representation through a one-dimensional convolutional neural network. The one-dimensional convolutional neural network (1D-CNN) is suitable for processing sequence data. By sliding the convolutional kernel in the time dimension, it captures local pattern changes. In the elevator state prediction, the number of convolutional kernels in the 1D-CNN layer is set to 64, and the size of the convolutional kernel is 3, which means that the data of 3 consecutive time points are considered each time. The convolution operation calculates the dot product between the convolutional kernel and the local area of the input data to extract local temporal features. The sliding convolutional kernel moves backward from the starting position of the sequence, and a convolution calculation is performed every time it moves one step to obtain a set of feature maps. The ReLU activation function is used to increase non-linearity and enhance the expressive power of the model. 1D-CNN can effectively extract local patterns in elevator operation data, such as short-term vibration change trends, temperature fluctuation patterns, etc., and generate feature map results.

[0084] Input the feature map results into the fully connected layer for feature fusion. After flattening the feature maps extracted by the previous layer, the fully connected layer is connected to all neurons through a weight matrix to achieve the fusion of different features. In the elevator state prediction model, two fully connected layers are set, with the number of nodes being 64 and 32 in sequence, and the ReLU activation function is used. At the same time, the dropout mechanism is introduced to control the risk of overfitting. Dropout is a regularization technique that randomly discards a part of the neurons during the training process to prevent the model from relying too much on certain features. In the elevator state prediction model, the dropout rate is set to 0.3, that is, 30% of the neurons are randomly discarded each time during training. The fully connected layer finally outputs the fusion index prediction value for the next 8 hours and the future values of each key parameter, forming a prediction output vector.

[0085] Calculate the mean squared error loss based on the predicted output vector and the true label. Mean squared error (MSE) is a commonly used loss function in regression problems, which calculates the average of the sum of the squares of the differences between the predicted values and the true values. In elevator status prediction, the difference between the future 8-hour fusion index predicted by the model and the actual observed value is taken, squared, and then averaged to obtain the loss value. Parameter updates are performed through the Adam optimizer, which combines the advantages of the momentum method and the adaptive learning rate method, and dynamically adjusts the learning rate according to the gradient. The initial learning rate is set to 0.001, and the learning rate adopts a decay strategy, with the learning rate decreasing by 10% every 50 epochs. To prevent overfitting, when the loss on the validation set has not decreased for 10 consecutive epochs, the early stopping mechanism is activated to stop the training. After the complete training process, the average prediction error of the elevator status prediction model on the test set does not exceed 3.5%, and the correlation coefficient is not less than 0.92, ensuring the reliability of the prediction results.

[0086] Taking the elevators in a certain residential community as an example, two years of operation data were collected, and about 15,000 training samples were obtained through sliding window division. Each sample contains 24 hours of input data (fusion index and 7 types of original parameters) and the corresponding 8-hour prediction target. The bidirectional LSTM layer processes the time series information, capturing the long-term correlation between elevator load and vibration, as well as the periodic pattern of temperature change. The attention mechanism analysis found that the data 4-6 hours before the occurrence of vibration anomalies is the most critical for prediction, and the attention weights at the corresponding time points are significantly higher than those at other time points. One-dimensional CNN extracts local feature patterns such as rapid increase in motor temperature and change in vibration frequency. The fully connected layer fuses these features to generate the status prediction for the next 8 hours. The model can accurately predict the vibration anomalies caused by elevator guide rail wear and issue a warning 6 hours in advance. When it is detected that the motor temperature continues to rise and the vibration frequency shows anomalies, the model predicts that the fusion index will drop from 0.75 to 0.45 in the next 4 hours, triggering a "warning" level warning. Maintenance personnel accordingly checked and replaced the worn parts in advance, avoiding the occurrence of elevator failures.

[0087] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0088] (1) Discretize the predicted fusion index value and the predicted data of each parameter output by the elevator status prediction model, and divide them according to the three-level standards of normal state, mild abnormal state, and severe abnormal state to obtain the discrete data of the parameter status;

[0089] (2) Establish a transaction database based on the discrete data of the parameter status and the historical fault records, and associate and record the parameter status combinations with the corresponding fault types to obtain the status-fault mapping data set;

[0090] (3) Set the minimum support threshold and minimum confidence threshold for the state-fault mapping data set, and screen out the frequently occurring single-parameter abnormal states through support calculation to obtain the frequent item set;

[0091] (4) Perform a self-join operation based on the frequent item set, and construct a multi-parameter combination frequent item set through iterative connection and pruning methods to obtain the parameter combination feature set;

[0092] (5) Extract association rules from the parameter combination feature set, and obtain the preliminary association rules between parameter anomalies and fault types by calculating three indicators: support, confidence, and lift;

[0093] (6) Introduce a time-series information weighting mechanism for the preliminary association rules, assign different weight values according to the proximity of the data time, and sort according to the weighted confidence and lift to obtain the association rule set between parameter anomaly combinations and fault types.

[0094] Specifically, discretize the fusion index prediction value and various parameter prediction data output by the elevator state prediction model. Discretization is the process of converting continuous numerical values into discrete states, which is convenient for subsequent association rule mining. For the fusion index, divide the value range [0, 1] into five levels: excellent [0.8, 1.0], good [0.6, 0.8), average [0.4, 0.6), poor [0.2, 0.4), and dangerous [0, 0.2). For each parameter data, according to the equipment operation standards and historical statistical results, divide the parameter values into three levels: normal state, slightly abnormal state, and severely abnormal state. For example, the motor temperature below 60°C is in the normal state, 60 - 75°C is in the slightly abnormal state, and above 75°C is in the severely abnormal state; the vibration acceleration less than 0.2g is in the normal state, 0.2 - 0.4g is in the slightly abnormal state, and greater than 0.4g is in the severely abnormal state. Through this discretization process, the continuous monitoring data of elevator operation is converted into discrete parameter state data for easy analysis.

[0095] Then, establish a transaction database based on the parameter state discrete data and historical fault records. The transaction database is the basis for association rule mining, and each record contains the parameter state combination and the corresponding fault type. For each historical fault case, extract the parameter state discrete data within a period of time before the fault (such as 24 hours before the fault), and associate it with the fault type. For example, a record may contain "{vibration increase = severely abnormal, motor temperature rise = slightly abnormal, noise increase = severely abnormal, fault type = guide rail fault}". By organizing the relevant data of all historical fault cases, a complete state-fault mapping data set is formed to provide data support for subsequent association rule mining.

[0096] After obtaining the status-fault mapping dataset, the improved Apriori algorithm is applied for association rule mining. The Apriori algorithm is a classical association rule mining algorithm used to discover frequently occurring item sets in a dataset and the association relationships between them. First, set the minimum support threshold and the minimum confidence threshold. Support represents the proportion of records containing a certain item set in the total records, and confidence represents the proportion of records containing item set X that also contain item set Y. In elevator fault diagnosis, the minimum support is usually set relatively low (such as 0.03) because some important faults may occur less frequently; the minimum confidence is set relatively high (such as 0.75) to ensure the reliability of the rules. By calculating the occurrence frequency of a single parameter abnormal state in the dataset, filter out the parameter states whose support is not lower than the minimum support to form frequent 1-item sets.

[0097] Based on the frequent 1-item sets, perform self-join operations to construct higher-order frequent item sets. Self-join is the process of joining a frequent k-item set with itself to generate candidate k+1-item sets. During the joining process, it is required that two k-item sets have k-1 items in common. For example, from the frequent 1-item sets {Vibration increase = Seriously abnormal} and {Motor temperature rise = Slightly abnormal}, the candidate 2-item set {Vibration increase = Seriously abnormal, Motor temperature rise = Slightly abnormal} can be generated. After generating the candidate item sets, pruning operations need to be performed to delete those candidate k-item sets whose any k-1 subsets are not frequent item sets. By iteratively performing the joining and pruning operations, gradually construct higher-order frequent item sets until no new frequent item sets can be generated. In this way, all the frequent item sets are obtained, forming a parameter combination feature set that contains the characteristics of various parameter abnormal combinations during elevator operation.

[0098] Extract association rules from the parameter combination feature set to form a description of the relationship between parameter abnormalities and fault types. For each frequent item set, enumerate all its possible non-empty proper subsets as the antecedent and the remaining part as the consequent to form candidate association rules. Then calculate the three indicators of support, confidence, and lift for each rule. Support represents the proportion of the rule in the total dataset; confidence represents the probability that the consequent occurs when the antecedent occurs; lift represents the ratio of the occurrence frequency of the consequent in the records containing the antecedent to the occurrence frequency of the consequent in the overall records, which is used to measure the effectiveness of the rule. A lift greater than 1 indicates that the occurrence of the antecedent will increase the possibility of the consequent occurring, and the rule has positive correlation. Filter out the rules that simultaneously meet the requirements of the minimum support and the minimum confidence to obtain the preliminary association rules between parameter abnormalities and fault types.

[0099] Finally, a time - series information weighting mechanism is introduced to the preliminary association rules to improve the timeliness of the rules. The operating state of the elevator is dynamically changing, and recent data has a stronger indication of the current state than long - term data. The time - series information weighting mechanism assigns different weights according to the proximity of the data generation time, with higher weights for recent data and lower weights for long - term data. The weight calculation uses an exponential decay function to give higher importance to recent data. The rules are re - sorted based on the weighted confidence and lift to form the final set of association rules between parameter anomaly combinations and fault types. These rules express the relationship of "if a certain parameter anomaly combination occurs, then a certain type of fault may occur", providing knowledge support for elevator fault diagnosis and early warning.

[0100] Taking the elevator in an office building as an example, by analyzing the operation data and fault records over the past six months, a transaction database containing 2000 records was established. The discretization process converts continuous monitoring data into state descriptions. For example, the vibration data of 0.45g is converted into "vibration increase = severe anomaly". The minimum support threshold is set to 0.03, and the minimum confidence threshold is set to 0.75. Frequent item sets are mined through the Apriori algorithm. For example, the frequent 1 - item sets include "vibration increase = severe anomaly" (support 0.06), "noise increase = severe anomaly" (support 0.05), etc. Through the self - joining and pruning of frequent item sets, multi - parameter combination frequent item sets are constructed, such as {vibration increase = severe anomaly, motor temperature increase = slight anomaly, noise increase = severe anomaly}. Association rules are extracted from them, such as "{vibration increase = severe anomaly, motor temperature increase = slight anomaly, noise increase = severe anomaly}→{guide rail fault}". The support of this rule is 0.045, the confidence is 0.86, and the lift is 12.3, indicating that this parameter anomaly combination is highly correlated with guide rail faults. Through time - series weighting processing, the weight of data within the recent two weeks is 1.5 times that of long - term data, further improving the timeliness of the rules. The final set of association rules can effectively guide elevator fault diagnosis. When a parameter anomaly combination that matches the antecedent of the rule is detected, the early warning platform will issue a corresponding type of fault warning in a timely manner.

[0101] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0102] (1) Construct an adaptive dynamic threshold system, and obtain the alarm threshold of elevator operation parameters through comprehensive calculation of the historical mean and standard deviation of parameters, elevator type, service life, and load status;

[0103] (2) According to the matching results of the set of association rules, through the comparative analysis of the parameter anomaly state and the antecedent of the rule, divide the elevator operation state into four levels: normal, attention, warning, and danger, and obtain hierarchical early warning information;

[0104] (3) Calculate the probability of each potential fault type based on the hierarchical early warning information, and obtain the fault probability distribution through the product operation of the confidence level, lift degree of the association rule and the matching degree of the current parameter state.

[0105] (4) Conduct trend analysis on the fault type data with a relatively high fault probability distribution, infer the fault development speed through the exponential smoothing prediction method, and obtain the estimated fault time result.

[0106] (5) Retrieve historical cases similar to the fault type from the fault knowledge base, extract relevant maintenance experience through the calculation of fault feature similarity, and obtain the reference maintenance plan.

[0107] (6) Integrate the hierarchical early warning information, fault probability distribution, estimated fault time result and reference maintenance plan into structured data to obtain the fault diagnosis report.

[0108] Specifically, an adaptive dynamic threshold system is constructed. Traditional fixed thresholds cannot adapt to the differences of different elevator types and different operating conditions, while the adaptive dynamic threshold can dynamically adjust the alarm boundary according to the characteristics of the elevator. The calculation of the adaptive threshold takes into account the historical mean value, standard deviation of the parameters and the specific situation of the elevator, and uses the formula: Threshold = Historical mean value of the parameter ± Adaptive coefficient × Historical standard deviation of the parameter. Among them, the adaptive coefficient is jointly determined by factors such as elevator type, service life, load status, etc., and the calculation method is: Adaptive coefficient = Basic coefficient × [1 + Service life influence factor × log (Service life + 1) + Load influence factor × Load rate + Frequency influence factor × Operating frequency]. The basic coefficient is set according to the elevator type and is usually between 2.5 and 3.5; the service life influence factor is about 0.15, and the longer the service life, the wider the threshold range; the load influence factor is about 0.1, and the higher the load rate, the higher the upper threshold; the frequency influence factor is about 0.08, and the higher the operating frequency, the greater the allowable threshold fluctuation. Through this dynamic calculation method, the alarm thresholds of various elevator operating parameters are obtained, providing a benchmark for state assessment.

[0109] Next, based on the matching results of the association rule set, a hierarchical evaluation of the elevator operation status is carried out. The elevator operation status is divided into four levels: normal, attention, warning, and danger. The division basis is the matching situation between the current parameter abnormal status and the antecedent of the rule. The normal status means that all parameters are within the normal range and no association rules are triggered; the attention status means that there are slightly abnormal parameters or association rules with low confidence are triggered; the warning status means that multiple parameters are abnormal or association rules with high confidence are triggered; the danger status means that key parameters are severely abnormal or association rules with extremely high confidence are triggered. The specific judgment process is to compare the status of the current elevator operation parameters with the antecedents of each rule in the association rule set and calculate the matching degree. The matching degree is the degree of coincidence between the current status and the antecedent of the rule. A complete coincidence is 1, and a partial coincidence is between 0 and 1. According to the highest matching degree and the confidence of the corresponding rule, the elevator operation status level is determined, and a hierarchical early warning information is formed.

[0110] Based on the hierarchical early warning information, the occurrence probability of each potential fault type is calculated. For each possible fault type, probability evaluation is carried out through all the rules related to this fault in the association rule set. The fault probability calculation formula is: fault probability = the sum of (confidence × lift × current status matching degree) of all relevant rules. The confidence reflects the reliability of the rule, the lift reflects the correlation strength of the rule, and the current status matching degree reflects the degree of coincidence between the current situation and the rule conditions. For example, for a rule "vibration increase + temperature increase → guide rail fault" with a confidence of 0.85, a lift of 12, and a matching degree of 0.9 between the current status and the antecedent of the rule, the contribution of this rule to the guide rail fault probability is 0.85×12×0.9 = 9.18. Add up the contributions of all the rules related to the guide rail fault to get the probability score of the guide rail fault. Calculate the probability scores of various faults in this way to form a fault probability distribution, indicating the most likely fault type currently.

[0111] Trend analysis is carried out on the fault types with higher values in the fault probability distribution to estimate the fault occurrence time. The exponential smoothing prediction method is adopted, which is suitable for processing time series data with trends. The exponential smoothing prediction assigns higher weights to recent data and lower weights to far - term data, and predicts the future trend through a weighted average method. The specific calculation process is to use the parameter deterioration trend to infer the time point when the parameter reaches the fault critical value. For example, by analyzing the motor temperature rise rate, predict the time required for the temperature to reach the fault critical value of 80°C; by analyzing the vibration amplitude growth trend, predict the time when the vibration reaches the danger threshold of 0.6g. Combine the prediction results of all relevant parameters to obtain the time range when the fault may occur, form the fault time estimation result, and provide a time reference for the maintenance plan.

[0112] Retrieve historical cases similar to the current warning fault type from the fault knowledge base to provide reference for maintenance. The fault knowledge base records historical fault cases, including information such as fault descriptions, parameter characteristics, fault causes, maintenance measures, and maintenance effects. By calculating the similarity between the current fault characteristics and historical cases, the most relevant cases are identified. The similarity calculation is based on the Euclidean distance or cosine similarity of fault parameter characteristics. The specific process is to compare the parameter feature vector of the current fault with the feature vectors of historical cases. For example, if the current fault characteristics include an increase in vibration frequency, an increase in motor temperature, and an increase in noise decibels, after representing them as feature vectors, the similarity is calculated with the cases in the knowledge base, and historical cases with a similarity higher than the threshold are selected. Maintenance experience is extracted from these similar cases, including information such as fault cause analysis, maintenance method suggestions, required tools and materials, etc., to form a reference maintenance plan.

[0113] Finally, the hierarchical warning information, fault probability distribution, fault time estimation results, and reference maintenance plan are integrated into structured data to generate a fault diagnosis report. The report includes the following parts: The basic information part records basic data such as elevator number, model, installation location, and service life; the overall operation status part shows statistical information such as recent operation duration, passenger capacity, and operation frequency; the abnormal parameter analysis part lists all parameters that exceed the normal range, including parameter name, normal range, current value, deviation degree, change trend, etc.; the fault diagnosis result part gives possible fault types, fault probabilities, fault cause analysis, etc.; the reference case part provides historical similar fault cases for reference; the maintenance suggestion part puts forward specific maintenance suggestions according to the fault type and severity, including maintenance items, priorities, estimated working hours, and required materials, etc.; the preventive measure part puts forward suggestions for long-term prevention of similar faults. The fault diagnosis report is managed by grading according to severity, and reports at different levels adopt different transmission methods and processing procedures to ensure that important information is processed in a timely manner.

[0114] Taking a high-rise residential elevator as an example, through multi-source data fusion monitoring, it is found that the vibration data of the elevator is abnormal. The acceleration value gradually rises from the normal 0.15g to 0.38g, while the noise increases to 72 decibels and the motor temperature rises to 68°C. Considering the characteristics that the elevator has been in use for 8 years and the average load rate is 60%, the adaptive threshold system calculates the vibration alarm threshold as 0.35g, the noise alarm threshold as 70 decibels, and the motor temperature alarm threshold as 65°C. According to the matching result of the association rule set, the matching degree of the current state with the rule "{Vibration increase = Slight abnormality, Noise increase = Slight abnormality, Motor temperature rise = Slight abnormality} → {Guide rail failure}" is 0.85, and the confidence level of this rule is 0.82, triggering a "warning" level early warning. Calculating the probabilities of various fault types, the probability of guide rail failure is the highest, at 78%, the probability of wire rope wear is 15%, and the probabilities of other faults are relatively low. Through exponential smoothing prediction, the vibration value will reach the dangerous threshold of 0.5g within 72 hours, and it is recommended to carry out maintenance within 72 hours. Three similar cases are retrieved from the fault knowledge base, all of which are vibration abnormalities caused by guide rail wear. The maintenance experience is extracted to form a reference maintenance plan, including specific measures such as checking whether the guide rail is deformed, checking whether the guide shoe is worn, and adjusting the guide shoe clearance. A structured fault diagnosis report is generated and submitted to the maintenance personnel for targeted maintenance, avoiding the occurrence of elevator failures.

[0115] In a specific embodiment, the process of extracting relevant maintenance experience through calculation of fault feature similarity can specifically include the following steps:

[0116] (1) Construct a feature vector representation for the fault type, and convert it into a structured representation through the abnormal mode of key parameters, the description of the equipment operation state, and the encoding of fault phenomena, to obtain the target fault feature vector;

[0117] (2) Extract historical fault case data from the fault knowledge base, and obtain the historical case feature set by structurally extracting the fault features, maintenance measures, and maintenance effects of each case;

[0118] (3) Calculate the similarity based on the target fault feature vector and the historical case feature set, and quantify the matching degree of the fault mode through the cosine similarity method to obtain the case similarity sorted list;

[0119] (4) Screen historical cases with similarity higher than the threshold from the case similarity sorted list, and ensure the relevance of the reference cases by setting the lowest limit of similarity to obtain relevant historical maintenance cases;

[0120] (5) Conduct cluster analysis on the maintenance measures in the relevant historical maintenance cases, group the maintenance methods according to the relevance of the fault causes and component positions, and obtain the maintenance plan categories;

[0121] (6) Based on the maintenance plan category and historical success rate data, by calculating the effectiveness index and resource consumption score of each plan, a priority list is formed through comprehensive ranking to obtain a reference maintenance plan.

[0122] Specifically, the multi-dimensional data collected by various sensors during the elevator operation need to undergo fine feature extraction and coding conversion. Specifically, the construction of feature vectors includes the structured characterization of key parameters such as the spectral characteristics of vibration signals, the motor temperature change curve, and the load fluctuation trend. For example, for traction system failures, the vibration signal will be decomposed into frequency components, amplitude characteristics, and time characteristics, and key indicators such as abnormal points and change slopes will be extracted from the motor temperature change. These data are converted into multi-dimensional feature vectors through standardized coding. The extraction of data from the fault knowledge base follows strict structured principles. Each historical fault case is regarded as a complete knowledge unit, and the extraction process not only collects basic fault information but also includes detailed maintenance technical routes, types of spare parts used, details of maintenance personnel operations, and the equipment performance recovery after maintenance. For elevator control system failures, the knowledge base will record specific fault diagnosis processes, types of control boards replaced, debugging parameters, fault duration, and other fine information to ensure that each case can provide comprehensive technical references.

[0123] The similarity calculation uses the cosine similarity algorithm as the core quantization method. This algorithm accurately evaluates the matching degree of fault patterns by calculating the cosine value of the angle between the target fault feature vector and the historical case feature set in the multi-dimensional feature space. The calculation process involves mapping the high-dimensional feature vector to a unified feature space and normalizing the features of each dimension to eliminate the influence of dimension and scale differences. For example, for brake system failures, the system compares the similarities of multiple dimensions such as abnormal vibration frequency, temperature change characteristics, and current fluctuations in the target fault feature vector with historical cases. Case screening achieves precise matching by setting a similarity threshold. The threshold is set based on expert experience and big data statistical analysis to ensure that only highly relevant historical maintenance cases are included in the reference range. The screening process not only considers the similarity score but also comprehensively analyzes factors such as the timeliness of the case and the matching degree of equipment types. For elevator traction system failures, the system may set the similarity threshold to 0.75 and only select historical maintenance cases with high feature vector matching degrees. Maintenance measure clustering analysis is professionally grouped based on the fault cause and component location. The clustering algorithm classifies historical maintenance cases in detail according to dimensions such as the fault location, maintenance technical route, and spare parts used. For the elevator brake system, maintenance plan categories such as electrical control, mechanical component replacement, lubrication and commissioning may be formed. Each category represents a typical maintenance processing logic, reflecting the professionalism and systematicness of elevator maintenance. The scheme priority ranking comprehensively considers the maintenance effectiveness index and resource consumption score. The effectiveness index is calculated based on indicators such as the maintenance success rate and equipment recovery time of historical cases; the resource consumption score includes multiple dimensions such as spare part cost, labor cost, and downtime. Through multi-dimensional weighted evaluation, a maintenance plan priority list is generated.

[0124] The above described the elevator operation status monitoring and early warning method based on multi-source data fusion in the embodiments of the present application. Next, the elevator operation status monitoring and early warning platform based on multi-source data fusion in the embodiments of the present application will be described. Please refer to Figure 2 , an embodiment of the elevator operation status monitoring and early warning platform based on multi-source data fusion in the embodiments of the present application includes:

[0125] A processing module, configured to collect and preprocess vibration signals, noise, speed fluctuations, motor temperature, load data, door machine parameters, and guide rail deformation data during the operation of the elevator to obtain normalized multi-source data;

[0126] A fusion module, configured to perform fusion processing on the normalized multi-source data based on the improved entropy weight method, and adjust the weight distribution by introducing a sensitivity correction factor to obtain an elevator operation status fusion index;

[0127] An input module for inputting the elevator operation status fusion index and the original multi-source data into a hybrid deep learning model composed of a bidirectional long short-term memory network, an attention mechanism, and a convolutional neural network for training to obtain an elevator status prediction model;

[0128] An analysis module for performing a correlation analysis on the parameter status and the fault type according to the output result of the elevator status prediction model to obtain a set of association rules between parameter abnormal combinations and fault types;

[0129] An evaluation module for evaluating the elevator operation status based on the set of association rules to obtain hierarchical warning information, and generating a fault diagnosis report based on the hierarchical warning information.

[0130] Through the collaborative cooperation of the above-mentioned various components, by innovatively integrating technical features such as multi-source sensor data acquisition, improved entropy weight data fusion, hybrid deep learning model construction, and intelligent fault association analysis, the scientific nature and accuracy of elevator operation status monitoring have been significantly improved. At the data acquisition level, through the collaborative work of various sensors such as acceleration sensors, sound sensors, and speed sensors, a comprehensive capture of multi-dimensional features such as vibration signals, noise, speed fluctuations, motor temperature, and load data during the elevator operation process has been achieved, laying a solid data foundation for subsequent in-depth analysis. The improved entropy weight method introduces a sensitivity correction factor, breaking through the limitations of the weight allocation of the traditional entropy weight method, and being able to adjust the weights of different data sources more dynamically and accurately, effectively balancing the importance of various sensor data. The innovation of the hybrid deep learning model lies in integrating a bidirectional long short-term memory network, an attention mechanism, and a convolutional neural network, giving full play to the advantages of artificial intelligence algorithms in processing complex time-series data. Specifically, the bidirectional LSTM network captures long-term dependence relationships, the attention mechanism realizes dynamic weight allocation for key time nodes, and the one-dimensional convolutional neural network extracts local feature patterns. The organic combination of the three algorithms significantly improves the accuracy and robustness of elevator operation status prediction. The introduction of association rule mining technology realizes a deep correlation analysis between parameter abnormal states and fault types, and constructs an intelligent early warning system for elevator faults by setting dynamic thresholds and time-series information weighting. The fault case retrieval and maintenance plan recommendation links adopt a feature vector similarity matching and multi-dimensional scoring mechanism, breaking through the traditional experience-driven maintenance mode, and establishing a data-driven scientific maintenance paradigm. This technical path of multi-source data fusion and artificial intelligence algorithm empowerment significantly improves the accuracy and early warning ability of elevator operation status monitoring.

[0131] Refer to Figure 3 In the embodiment of the present invention, a computer device is further provided. The computer device may be a server, and its internal structure may be as Figure 3As shown in the figure. The computer device includes a processor, a memory, a display screen, an input device, a network interface, and a database connected through a system bus. Among them, the processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. The computer program, when executed by the processor, implements the above method.

[0132] Those skilled in the art can understand that Figure 3 the structure shown in the figure is only a block diagram of some structures related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0133] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0134] Those of ordinary skill in the art can understand that all or part of the process of implementing the method in the above embodiment can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium provided by the present invention and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or an external cache. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0135] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the systems, systems, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0136] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0137] The above is the case. The above embodiments are only used to illustrate the technical solutions of the present application and are not intended to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments or perform equivalent replacements for some of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present application.

Claims

1. An elevator operation status monitoring and early warning method based on multi-source data fusion, characterized in that, The elevator operation status monitoring and early warning method based on multi-source data fusion includes: Collect and preprocess vibration signals, noise, speed fluctuations, motor temperature, load data, door machine parameters, and guide rail deformation data during the elevator operation to obtain normalized multi-source data; Based on the improved entropy weight method, fuse and process the normalized multi-source data, and adjust the weight distribution by introducing a sensitivity correction factor to obtain the elevator operation status fusion index; Input the elevator operation status fusion index and the original multi-source data into a hybrid deep learning model composed of a bidirectional long short-term memory network, an attention mechanism, and a convolutional neural network for training to obtain an elevator status prediction model; According to the output result of the elevator status prediction model, conduct a correlation analysis on the parameter status and fault types to obtain a set of association rules for parameter anomaly combinations and fault types; Based on the set of association rules, evaluate the elevator operation status to obtain hierarchical early warning information, and generate a fault diagnosis report based on the hierarchical early warning information.

2. The elevator operation status monitoring and early warning method based on multi-source data fusion according to claim 1, characterized in that The step of collecting and preprocessing vibration signals, noise, speed fluctuations, motor temperature, load data, door machine parameters, and guide rail deformation data during the elevator operation to obtain normalized multi-source data includes: Collect original operation data through various sensor devices set in the elevator system, including collecting vibration signals through acceleration sensors, collecting noise data through sound sensors, collecting speed fluctuation values through speed sensors, collecting motor temperature through temperature sensors, collecting load data through load sensors, collecting door machine parameters through the door machine control system, and collecting guide rail deformation data through displacement sensors, to obtain the original elevator operation monitoring data; Eliminate environmental interference from the original elevator operation monitoring data through the wavelet transform method, and separate the effective signal and noise according to the directional noise reduction algorithm to obtain the elevator operation noise-reduced data; Based on the elevator operation noise-reduced data, supplement the missing data through linear interpolation or forward filling method, and fill the data blank points according to the principle of time series continuity to obtain the elevator operation complete data with a completeness of not less than 99.5%; Synchronize the elevator operation complete data through the timestamp alignment function, and reconstruct the data streams of each sensor according to a unified sampling period to obtain the elevator monitoring data with time series alignment; Identify and replace the outlier points in the elevator monitoring data with time series alignment using the 3σ principle, and screen out the data points deviating from the normal range based on the statistical distribution characteristics to obtain the elevator monitoring data after outlier processing; Convert the elevator monitoring data after outlier processing to the [-1, 1] interval according to the maximum-minimum normalization method to eliminate the influence of different physical dimensions and obtain the normalized multi-source data.

3. The elevator operation status monitoring and early warning method based on multi-source data fusion according to claim 1, wherein, The step of fusing and processing the normalized multi-source data based on the improved entropy weight method, adjusting the weight distribution by introducing a sensitivity correction factor to obtain the elevator operation status fusion index includes: Construct a standardized data matrix for the normalized multi-source data, and obtain a standardized processing matrix through the normalization calculation of positive indicators and negative indicators; Calculate the proportion and information entropy of each index at each sampling point according to the standardized processing matrix, and obtain the information entropy of each index through the operation of multiplying the proportion by the logarithm and then summing up; Calculate the coefficient of variation and the initial weight based on the information entropy of each index. Obtain the coefficient of variation by subtracting the information entropy from the constant 1, and obtain the initial weight by dividing the coefficient of variation by the sum, so as to get the initial weight configuration; Calculate the sensitivity correction factor through the ratio of the standard deviation of each index to the maximum standard deviation, and use the ratio plus 1 as the correction factor to obtain the weight adjustment parameter; Use the weight adjustment parameter to correct the initial weight, and obtain the final weight by multiplying the correction factor by the initial weight and normalizing; Based on the weighted summation of the final weight and the standardized data, establish a sub-scenario evaluation system for elevator operation under different working conditions, and obtain the elevator operation state fusion index; 4. The elevator operation status monitoring and early warning method based on multi-source data fusion according to claim 1, characterized in that Input the elevator operation state fusion index and the original multi-source data into a hybrid deep learning model composed of a bidirectional long short-term memory network, an attention mechanism and a convolutional neural network for training, and obtain an elevator state prediction model, including: Divide the historical time series data by a sliding window, set an input duration of 24 hours and a prediction duration of 8 hours to form a training sample set; Input the elevator operation state fusion index and the original multi-source data in the training sample set into the bidirectional long short-term memory network layer, and capture long-term dependence relationships through forward propagation and backward propagation to obtain sequence feature representations; Construct an attention weight coefficient based on the sequence feature representation, and calculate the importance distribution of each time point through the softmax function to obtain a weighted feature representation; Perform local feature extraction on the weighted feature representation through a one-dimensional convolutional neural network, and use a sliding convolutional kernel to capture local pattern changes to obtain a feature mapping result; Input the feature mapping result into a fully connected layer for feature fusion, and at the same time introduce a dropout mechanism to control the risk of overfitting to obtain a prediction output vector; Calculate the mean square error loss according to the prediction output vector and the true label, update the parameters through the Adam optimizer, and stop training when the loss on the validation set no longer decreases to obtain the elevator state prediction model; 5. The elevator operation status monitoring and early warning method based on multi-source data fusion according to claim 1, characterized in that According to the output result of the elevator state prediction model, conduct a correlation analysis on the parameter state and the fault type to obtain a set of association rules between parameter anomaly combinations and fault types, including: Discretize the predicted value of the fusion index and the predicted data of each parameter output by the elevator state prediction model, and divide them according to the three-level standards of normal state, slight anomaly state and serious anomaly state to obtain discrete data of the parameter state; Establish a transaction database according to the discrete data of the parameter state and the historical fault records, and associate the parameter state combinations with the corresponding fault types to obtain a state-fault mapping data set; Set a minimum support threshold and a minimum confidence threshold for the state-fault mapping data set, and screen out the single parameter anomaly states that frequently appear through support calculation to obtain frequent item sets; Perform a self-join operation on the frequent item sets, and construct a multi-parameter combination frequent item set through iterative connection and pruning to obtain a parameter combination feature set; Extract association rules from the parameter combination feature set, and obtain the preliminary association rules between parameter anomalies and fault types by calculating three indicators: support, confidence, and lift; Introduce a time-series information weighting mechanism for the preliminary association rules, assign different weight values according to the distance of data time, and sort according to weighted confidence and lift to obtain the association rule set between the parameter anomaly combination and the fault type; 6. The elevator operation status monitoring and early warning method based on multi-source data fusion according to claim 5, characterized in that Evaluate the elevator operation status based on the association rule set to obtain hierarchical early warning information, and generate a fault diagnosis report based on the hierarchical early warning information, including: Construct an adaptive dynamic threshold system, and obtain the alarm threshold of elevator operation parameters through comprehensive calculation of the historical mean and standard deviation of parameters, elevator type, service life, and load status; According to the matching results of the association rule set, divide the elevator operation status into four levels: normal, attention, warning, and danger through comparative analysis of the parameter anomaly status and the antecedent of the rule, and obtain the hierarchical early warning information; Calculate the probability of each potential fault type based on the hierarchical early warning information, and obtain the fault probability distribution through the product operation of the confidence and lift of the association rule and the matching degree of the current parameter status; Conduct trend analysis on the fault type data with a higher fault probability distribution, and infer the fault development speed through the exponential smoothing prediction method to obtain the estimated fault time result; Retrieve historical cases similar to the fault type from the fault knowledge base, and extract relevant maintenance experience through calculation of fault feature similarity to obtain a reference maintenance plan; Integrate the hierarchical early warning information, fault probability distribution, estimated fault time result, and reference maintenance plan into structured data to obtain the fault diagnosis report; 7. The elevator operation status monitoring and early warning method based on multi-source data fusion according to claim 6, characterized in that The retrieving historical cases similar to the fault type from the fault knowledge base, and extracting relevant maintenance experience through calculation of fault feature similarity to obtain a reference maintenance plan, includes: Construct a feature vector representation for the fault type, and convert it into a structured representation through the key parameter anomaly pattern, equipment operation status description, and fault phenomenon coding to obtain the target fault feature vector; Extract historical fault case data from the fault knowledge base, and obtain the historical case feature set by structurally extracting the fault features, maintenance measures, and maintenance effects of each case; Calculate the similarity between the target fault feature vector and the historical case feature set, and quantify the matching degree of the fault pattern through the cosine similarity method to obtain a list of case similarity rankings; Screen historical cases with a similarity higher than the threshold from the list of case similarity rankings, and ensure the relevance of the reference cases by setting the lowest limit of similarity to obtain relevant historical maintenance cases; Conduct cluster analysis on the maintenance measures in the relevant historical maintenance cases, and group the maintenance methods according to the relevance of the fault cause and component location to obtain the maintenance plan category; Based on the maintenance plan category and historical success rate data, calculate the effectiveness index and resource consumption score of each plan, and comprehensively sort to form a priority list to obtain the reference maintenance plan; 8. An elevator operation status monitoring and early warning platform based on multi-source data fusion is used to implement the elevator operation status monitoring and early warning method based on multi-source data fusion as described in any one of claims 1-7, and is characterized in that, The elevator operation status monitoring and early warning platform based on multi-source data fusion includes: A processing module, configured to collect and preprocess vibration signals, noise, speed fluctuations, motor temperature, load data, door machine parameters, and guide rail deformation data during the operation of the elevator to obtain normalized multi-source data; A fusion module, configured to perform fusion processing on the normalized multi-source data based on an improved entropy weight method, and adjust the weight allocation by introducing a sensitivity correction factor to obtain an elevator operation state fusion index; An input module, configured to input the elevator operation state fusion index and the original multi-source data into a hybrid deep learning model composed of a bidirectional long short-term memory network, an attention mechanism, and a convolutional neural network for training to obtain an elevator state prediction model; An analysis module, configured to perform a correlation analysis on the parameter state and the fault type according to the output result of the elevator state prediction model to obtain a set of association rules between parameter abnormal combinations and fault types; An evaluation module, configured to evaluate the elevator operation state based on the set of association rules to obtain hierarchical early warning information, and generate a fault diagnosis report based on the hierarchical early warning information.

9. A computer device, characterized in that, It includes a memory and a processor. The memory stores a computer program that can run on the processor. It is characterized in that when the processor executes the computer program, it implements the elevator operation state monitoring and early warning method based on multi-source data fusion according to any one of claims 1 to 7.

10. A computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, the processor is caused to execute the elevator operation state monitoring and early warning method based on multi-source data fusion according to any one of claims 1 to 7.

Images