An elevator counterweight running fault diagnosis system and method integrating vibration and sound information

Abstract

The present application relates to a kind of comprehensive vibration and sound information's elevator counter pulley operation fault diagnosis system and method, belong to elevator fault diagnosis technical field.The system, including signal acquisition layer, edge gateway layer, cloud analysis layer and client layer;Signal acquisition layer is synchronously collected counter pulley original signal by vibration and sound sensor;After edge gateway layer is preprocessed and framed alignment to signal, calculate vibration kurtosis and sound short-time energy to carry out double-threshold preliminary screening, only upload abnormal data section;Cloud extracts vibration characteristics such as frequency conversion, bearing fault frequency, narrowband peak value, sensitive subband energy, and sound characteristics such as MFCC, pitch frequency and its stability, after fusion, input random forest model, output fault type and health grade;Client is thus graded early warning.The present application realizes the accurate discrimination to bearing damage, steel wire rope off slot, wheel body crack and jam, reduces false alarm rate and transmission cost.

Description

Technical Field

[0001] This invention belongs to the field of elevator fault diagnosis technology, and specifically relates to an elevator anti-corrosion sheave operation fault diagnosis system and method that integrates vibration and sound information. Background Technology

[0002] The deflector pulley is a key guiding component in the elevator traction system, playing a crucial role in adjusting the output power and torque of the traction machine and changing the direction of the wire rope. Typically installed on the counterweight side or car side of the elevator car, the deflector pulley operates in a closed environment, enduring alternating tension, bending stress, and frictional loads from the wire rope over extended periods. During continuous operation, the deflector pulley is highly susceptible to mechanical failures such as bearing wear or jamming, fatigue cracks or spalling of the pulley body, and wire rope derailment. Because the deflector pulley is located inside the hoistway, the development of these failures is often concealed and difficult to detect. If jamming or breakage occurs, it can lead to serious safety accidents such as wire rope breakage and counterweight collapse.

[0003] Currently, the monitoring technologies for elevator anti-corrosion pulleys can be mainly divided into the following categories:

[0004] One existing technology, such as patent document CN121698199A, discloses a method for predicting the life of a non-metallic anti-cord sheave in elevators. This method involves assembling a triaxial vibration sensor, a temperature sensor, and strain gauges in the bearing housing to collect and fuse multi-source data, and then using an ε-SVR regression model for life prediction. This method focuses on long-term prediction of remaining life and requires a large amount of historical life data to train the regression model. Furthermore, the dynamic response of temperature and stress sensors to sudden mechanical jamming or instantaneous wire rope derailment exhibits lag, making real-time online diagnosis difficult.

[0005] The second existing technology, such as patent document CN121609181A, discloses a method for monitoring the failure of a reverse rope pulley based on image recognition technology. This method involves capturing video of the reverse rope pulley assembly using a camera, extracting image features of the wire rope, casing, and anti-slip components, and comparing these features with a standard image. However, this method can only extract macroscopic relative changes on the visual surface, making it difficult to detect initial microscopic wear vibrations and abnormal noises inside the bearing. Furthermore, visual monitoring is highly susceptible to interference from dust, oil, light, and obstructions from the installation angle within the shaft.

[0006] The third existing technology, such as patent document CN116101864A, discloses a fault diagnosis method for elevator door systems based on sound recognition technology, which diagnoses door system faults through MFCC features and a fault sound classifier. However, this method is only applied to elevator door systems, does not involve anti-cord pulley components, and does not combine vibration modes for mutual verification of two physical quantities.

[0007] In summary, existing technologies have the following shortcomings: they rely on a single-modal sensor, lack cross-verification of vibration and sound, and are prone to false alarms or missed alarms; they only focus on long-term lifespan prediction or macroscopic displacement changes, and lack the real-time diagnostic capability for microscopic mechanical faults such as bearing wear and wheel cracks inside the anti-rope reel; and all massive amounts of stable operation data are uploaded to the cloud, resulting in wasted bandwidth and storage. Therefore, a fault diagnosis solution for anti-rope reel operations that can integrate dual-modal information, possess preliminary edge-end diagnostic capabilities, and accurately distinguish multiple fault types is needed. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of the prior art and provide an elevator anti-cord sheave operation fault diagnosis system and method that integrates vibration and sound information.

[0009] To achieve the above objectives, the technical solution of the present invention is: an elevator anti-corrosion sheave operation fault diagnosis system that integrates vibration and sound information, comprising a signal acquisition layer, an edge gateway layer, a cloud analysis layer, and a client layer;

[0010] The signal acquisition layer includes a vibration sensor installed on the elevator anti-cord pulley bearing seat or pulley support structure, and a sound sensor installed on the anti-cord pulley guard or the wire rope swing area, for synchronously acquiring the original vibration signal and the original ambient sound signal during elevator operation.

[0011] The edge gateway layer is connected to the signal acquisition layer and is used to preprocess, frame, and align the original vibration signal and the original ambient sound signal in the time domain, and to filter out abnormal data segments by calculating the time domain kurtosis value of the vibration signal and the short-time energy value of the sound signal.

[0012] The cloud analysis layer is communicatively connected to the edge gateway layer and is used to receive the abnormal data segment, extract vibration side features and sound side features and fuse them, and output the fault type and health level through a classification model.

[0013] The client layer is connected to the cloud analysis layer and is used to obtain and display the fault type and health level.

[0014] Furthermore, the preprocessing performed by the edge gateway layer includes sequential bandpass filtering, wavelet threshold denoising, and DC component removal; the wavelet threshold denoising uses a soft threshold function, and the wavelet threshold is adaptively determined based on the median absolute difference of the wavelet coefficients and the signal length.

[0015] Furthermore, the vibration sensor is a triaxial accelerometer configured to capture radial and axial impact pulses from the bearing.

[0016] Furthermore, the cloud-based analysis layer includes a feature extraction and fusion module, which is configured to perform the following operations:

[0017] Fast Fourier Transform is performed on the abnormal vibration data segment to estimate the rotational frequency, and the bearing outer ring failure frequency BPFO and bearing inner ring failure frequency BPFI are calculated based on the number of rolling elements, rolling element diameter, pitch circle diameter, contact angle and estimated rotational frequency.

[0018] Search for spectral peaks at BPFO and BPFI frequencies to obtain narrowband search peaks for BPFO and BPFI.

[0019] The energy of the top m sensitive subbands with the largest energy percentage is extracted by wavelet packet decomposition.

[0020] Furthermore, the feature extraction and fusion module is also configured as follows:

[0021] Extract Mel frequency cepstral coefficients from abnormal sound data segments and calculate the mean and standard deviation of Mel frequency cepstral coefficients for each order.

[0022] Autocorrelation analysis or cepstral analysis is performed on the audio data segments to extract the fundamental frequency, and the standard deviation of the fundamental frequency in the abnormal data segments is calculated as a stability index.

[0023] Furthermore, the feature extraction and fusion module fuses the vibration-side features and the sound-side features to construct a multi-dimensional feature vector. The constructed multi-dimensional feature vector is composed of the following concatenated data:

[0024] Estimate the frequency shift, BPFO narrowband search peak, BPFI narrowband search peak, energy of each sensitive subband, mean cepstral coefficients of each Mel frequency, standard deviation of cepstral coefficients of each Mel frequency, mean and standard deviation of fundamental frequency.

[0025] Furthermore, the cloud-based analysis layer also includes a diagnostic classification module, which has a built-in random forest classification model. The random forest classification model is configured to output the corresponding fault type and health level based on the input multidimensional feature vector.

[0026] This invention also provides a method for diagnosing elevator anti-cord sheave operation faults by integrating vibration and sound information, based on the system described above, including:

[0027] Simultaneously collect raw vibration signals and raw ambient sound signals at the elevator anti-rope sheave;

[0028] The original vibration signal and the original ambient sound signal are sequentially subjected to bandpass filtering, wavelet threshold denoising, and DC component removal to obtain the denoised signal.

[0029] The denoised signal is framed and time-domain aligned according to the timestamp;

[0030] Calculate the time-domain kurtosis value of the vibration signal and the short-time energy value of the sound signal. If the kurtosis value is greater than the preset kurtosis threshold and the short-time energy value is greater than the preset energy threshold, then extract the corresponding abnormal vibration data segment and abnormal sound data segment.

[0031] Upload the abnormal vibration data segment and the abnormal sound data segment to the cloud;

[0032] The FFT and wavelet packet decomposition algorithms are executed on the abnormal vibration data segments in the cloud to extract the narrowband search peaks and sensitive subband energy at the frequency of the frequency conversion, BPFO and BPFI.

[0033] The mean and standard deviation of the Mel frequency cepstral coefficients are extracted from the abnormal sound data segments in the cloud, and the fundamental frequency and its stability index are also extracted.

[0034] Vibration-side features and sound-side features are fused into a multi-dimensional feature vector, which is then input into a random forest classification model to output the fault type and health level.

[0035] The client executes a tiered warning display based on the health level.

[0036] Furthermore, the judgment logic of the random forest classification model includes:

[0037] When a narrow-band search peak appears at BPFO, the vibration kurtosis value is greater than the preset threshold, the mean value of the high-frequency component of the Mel frequency cepstral coefficient is out of range and the fundamental frequency matches BPFO, it is determined to be bearing outer ring wear or pitting.

[0038] When a narrow-band search peak appears at BPFI, the vibration kurtosis value is greater than the preset threshold, the mean value of the Mel frequency cepstral coefficients shifts and the fundamental frequency matches BPFI, it is determined to be bearing inner ring wear or pitting.

[0039] When the vibration kurtosis value is greater than the preset threshold, the sound energy is concentrated in the mid-low frequency band for a short time, the standard deviation of the Mel frequency cepstral coefficient increases, there are no narrow band peaks at BPFO and BPFI and the fundamental frequency is unstable or cannot be extracted, it is determined to be a wire rope derailment and swinging.

[0040] When the energy of the high-frequency sensitive subband extracted by wavelet packet decomposition increases and the mean value of the Mel frequency cepstral coefficient is abnormal in the high-frequency region but the kurtosis value does not exceed the standard, it is judged as a crack or spalling of the anti-rope wheel body.

[0041] When the vibration kurtosis value continuously exceeds the standard, the short-term energy remains high in the continuous time window, the BPFO or BPFI frequency is not matched, the overall energy of each subband is too high and the fundamental frequency is disordered or disappears, it is judged as bearing jamming.

[0042] Furthermore, the tiered early warning is executed based on the health level: when the health level is normal, the client displays a green light; when the health level is alert, the client displays a yellow light and prompts for enhanced monitoring; when the health level is warning, the client displays a red light and prompts for immediate maintenance.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. Dual-modal complementary mechanism: The vibration and sound dual-modal signals are coupled and introduced into the fault diagnosis of the anti-rope wheel. The physical mechanisms of the two modes are independent and mutually corroborative, which greatly reduces the monitoring blind spots and false alarm rates caused by a single sensor.

[0045] 2. Cross-modal same-frequency verification mechanism: The fundamental frequency of sound and its stability are introduced to form cross-modal same-frequency verification with the frequency conversion, BPFO and BPFI extracted from the vibration side, which significantly improves the diagnostic confidence.

[0046] 3. Edge-cloud collaboration and efficient filtering: The edge gateway calculates kurtosis and short-term energy values ​​in real time and sets dual threshold triggering logic to filter out a large amount of redundant data during stable operation. Only high-value data segments suspected of being faulty are uploaded to the cloud, significantly reducing transmission bandwidth and cloud storage pressure.

[0047] 4. Accurate classification and interpretability: The cloud-based system integrates classic mechanical physical features (bearing failure frequency, wavelet packet energy) with sound statistical features (MFCC, fundamental frequency), and combines them with a random forest classification model to accurately distinguish five types of faults: bearing outer ring wear / pitting, bearing inner ring wear / pitting, wire rope degrooving and swinging, wheel body cracks / stripping, and bearing jamming, providing maintenance personnel with clear repair directions. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the overall system architecture provided in an embodiment of the present invention;

[0049] Figure 2 This is a schematic diagram of the installation position of the anti-rope wheel sensor provided in an embodiment of the present invention.

[0050] Figure 3 This is a schematic diagram of the preliminary diagnostic logic for an edge gateway provided in an embodiment of the present invention;

[0051] Figure 4 A flowchart of the cloud-based diagnostic algorithm provided in an embodiment of the present invention. Detailed Implementation

[0052] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] This invention provides an elevator anti-corrosion sheave operation fault diagnosis system that integrates vibration and sound information, including a signal acquisition layer, an edge gateway layer, a cloud analysis layer, and a client layer;

[0054] The signal acquisition layer includes a vibration sensor installed on the elevator anti-cord pulley bearing seat or pulley support structure, and a sound sensor installed on the anti-cord pulley guard or the wire rope swing area, for synchronously acquiring the original vibration signal and the original ambient sound signal during elevator operation.

[0055] The edge gateway layer is connected to the signal acquisition layer and is used to preprocess, frame, and align the original vibration signal and the original ambient sound signal in the time domain, and to filter out abnormal data segments by calculating the time domain kurtosis value of the vibration signal and the short-time energy value of the sound signal.

[0056] The cloud analysis layer is communicatively connected to the edge gateway layer and is used to receive the abnormal data segment, extract vibration side features and sound side features and fuse them, and output the fault type and health level through a classification model.

[0057] The client layer is connected to the cloud analysis layer and is used to obtain and display the fault type and health level.

[0058] This invention also provides a method for diagnosing elevator anti-cord sheave operation faults by integrating vibration and sound information, based on the system described above, including:

[0059] Simultaneously collect raw vibration signals and raw ambient sound signals at the elevator anti-rope sheave;

[0060] The original vibration signal and the original ambient sound signal are sequentially subjected to bandpass filtering, wavelet threshold denoising, and DC component removal to obtain the denoised signal.

[0061] The denoised signal is framed and time-domain aligned according to the timestamp;

[0062] Calculate the time-domain kurtosis value of the vibration signal and the short-time energy value of the sound signal. If the kurtosis value is greater than the preset kurtosis threshold and the short-time energy value is greater than the preset energy threshold, then extract the corresponding abnormal vibration data segment and abnormal sound data segment.

[0063] Upload the abnormal vibration data segment and the abnormal sound data segment to the cloud;

[0064] The FFT and wavelet packet decomposition algorithms are executed on the abnormal vibration data segments in the cloud to extract the narrowband search peaks and sensitive subband energy at the frequency of the frequency conversion, BPFO and BPFI.

[0065] The mean and standard deviation of the Mel frequency cepstral coefficients are extracted from the abnormal sound data segments in the cloud, and the fundamental frequency and its stability index are also extracted.

[0066] Vibration-side features and sound-side features are fused into a multi-dimensional feature vector, which is then input into a random forest classification model to output the fault type and health level.

[0067] The client executes a tiered warning display based on the health level.

[0068] The following is a detailed implementation process of the present invention.

[0069] Please see Figure 1 As shown in the figure, this embodiment provides an elevator anti-rope wheel operation fault diagnosis system that integrates vibration and sound information. The system is an end-to-cloud collaborative architecture, consisting of a signal acquisition layer, an edge gateway layer, a cloud analysis layer, and a client layer.

[0070] The signal acquisition layer includes a vibration sensor installed on the elevator anti-cord pulley bearing seat or pulley support structure, and a sound sensor installed on the anti-cord pulley guard or the wire rope swing area, for synchronously acquiring the original vibration signal and the original ambient sound signal during elevator operation.

[0071] The edge gateway layer receives the raw signals sent by the vibration sensor and the sound sensor, performs signal preprocessing on the raw vibration signal and the raw ambient sound signal, and then frames the two preprocessed signals and aligns them in the time domain according to the timestamp to obtain corresponding vibration signal frames and sound signal frames. The edge gateway layer calculates the temporal kurtosis value of the vibration signal and the short-time energy value of the sound signal. If the kurtosis value exceeds a preset kurtosis threshold and the short-time energy value exceeds a preset energy threshold, the captured abnormal vibration and sound data segment is uploaded to the cloud analysis layer; otherwise, the data segment is discarded or only the statistical features are uploaded, without uploading all the raw data.

[0072] The cloud-based analysis layer includes a feature extraction and fusion module and a diagnostic classification module. The feature extraction and fusion module is used to extract the rotational frequency, bearing fault frequency, narrowband search peak, and sensitive subband energy from the abnormal vibration data segment, and to extract the mean and standard deviation of the Mel frequency cepstral coefficients, fundamental frequency, and their stability from the abnormal sound data segment. The vibration-side features and sound-side features are fused to construct a multidimensional feature vector. The diagnostic classification module has a built-in trained random forest classification model, which is used to input the multidimensional feature vector and output the corresponding fault type and health level.

[0073] The client layer is used to obtain the fault type and health level from the cloud and display and issue warnings.

[0074] The fault diagnosis method for the system includes the following steps:

[0075] Step 1: Sensor Deployment and Synchronous Data Acquisition

[0076] like Figure 2 As shown, a triaxial acceleration vibration sensor 4 is press-fitted into the bearing housing of the elevator deflector 5 to capture radial and axial impact pulse signals of the bearing; a directional sound sensor 2 is installed on the deflector guard 3 or in the swing path area of ​​the wire rope 1 to collect the sound of the wire rope movement and abnormal noise of the wheel rotation. During elevator operation, the two sensors synchronously collect the original vibration signal and the original ambient sound signal, and send them to the edge gateway 6 deployed in the elevator machine room or shaft control cabinet via signal line 7.

[0077] Step 2: Edge Gateway Signal Preprocessing

[0078] After receiving the two raw signals, the edge gateway performs the following preprocessing operations in sequence:

[0079] Bandpass filtering: The passband range is set according to the operating conditions of the anti-rope pulley to filter out low-frequency structural vibration interference and high-frequency electromagnetic noise;

[0080] Wavelet thresholding for denoising: A soft thresholding function is used for denoising. The calculation formula is:

[0081]

[0082] in, A robust estimate of the noise standard deviation is provided (using the median absolute difference of wavelet coefficients). N is the signal length. Coefficients below a threshold are shrunk, while coefficients above the threshold are retained. Finally, the denoised signal is reconstructed.

[0083] DC component removal: Subtract the mean value from the denoised signal to eliminate the influence of the sensor's DC bias.

[0084] Step 3: Frame segmentation and temporal alignment

[0085] The preprocessed vibration and sound signals are divided into frames according to preset duration and step size, and the frames of the two signals are matched and aligned according to the absolute timestamp at the time of acquisition to ensure that the kurtosis frames and energy frames calculated later correspond to the same working condition instant.

[0086] Step 4: Preliminary diagnosis of dual thresholds for edge gateway

[0087] Reference Figure 3 The edge gateway logic shown calculates feature indices for the frame-aligned signals.

[0088] The formula for calculating the kurtosis value K of the vibration signal is:

[0089]

[0090] in, For the first frame after segmentation Vibration amplitude at each sampling point The amplitude mean of the frame is given, and N is the number of sampling points per frame. During normal operation, the vibration signal is approximately normally distributed, and the kurtosis value is stable around 3. When cracks or pitting occur in the bearing, the periodic impact pulses significantly increase the kurtosis value.

[0091] Short-time energy value of sound signal The calculation formula is:

[0092]

[0093] in, For the first frame after segmentation The sound amplitude at each sampling point. During normal operation, the ambient sound energy is relatively stable; when the wire rope derails or the wheel experiences abnormal friction, the energy in the corresponding frequency band changes abruptly.

[0094] Edge gateway preset kurtosis threshold and energy threshold When both conditions are met and If a preliminary anomaly is detected in the current time window, the vibration and sound data for that segment are extracted and uploaded to the cloud; otherwise, the data segment is discarded or only the statistical features are uploaded, without uploading all the original data.

[0095] Step 5: Data Upload

[0096] The edge gateway will send vibration and sound data segments that are initially identified as abnormal to the cloud analysis layer via 4G / 5G or wired network.

[0097] Step 6: Cloud Feature Extraction

[0098] Reference Figure 4 The cloud processing flow shown in the diagram extracts features from both vibration and sound signals in the cloud.

[0099] Regarding the vibration signal, a Fast Fourier Transform (FFT) is first performed to estimate the current rotational frequency of the anti-rope wheel through spectral peak search. Then, the bearing failure characteristic frequencies are calculated. The outer ring failure frequency (BPFO) and the inner ring failure frequency (BPFI) are as follows:

[0100]

[0101]

[0102] Where n is the number of rolling elements, d is the diameter of the rolling elements, and D is the pitch circle diameter. The contact angle is determined by searching for peak values ​​within narrow bands centered at BPFO and BPFI in the frequency spectrum. and .

[0103] Next, wavelet packet decomposition is used to decompose the vibration signal into L layers, total The energy of the i-th subband. for:

[0104]

[0105] in Let M be the wavelet packet coefficients, and M be the coefficient length. The first m subbands are selected as the sensitive subband energy features, sorted by energy percentage.

[0106] For the audio signal, Mel-frequency cepstral coefficients (MFCCs) are extracted. The conversion between Mel frequency and linear frequency f is as follows:

[0107]

[0108] After filtering by the Mel filter bank and taking the logarithmic energy, the coefficients of each MFCC are obtained through discrete cosine transform. The MFCC coefficients of all frames within this anomalous data segment are statistically analyzed, and the mean value of each order of MFCC is calculated. and standard deviation :

[0109]

[0110] Where T is the number of frames contained in the abnormal data segment.

[0111] Simultaneously, the fundamental frequency of the sound signal is extracted through autocorrelation analysis or cepstral analysis. And calculate its standard deviation within the outlier data segment. As a stability indicator of the fundamental frequency, the fundamental frequency reflects the periodic characteristics of sound. When there is periodic damage to the rotating parts of the anti-rope pulley, the fundamental frequency of the sound will correspond to the fault frequency extracted from the vibration side.

[0112] Step 7: Feature Fusion

[0113] The vibration-side features and the sound-side features are concatenated to form a multidimensional feature vector F:

[0114]

[0115] in The mean of the fundamental frequency. The standard deviation of the fundamental frequency.

[0116] Step 8: Random Forest Classification

[0117] The feature vector F is input into the trained random forest classification model, which outputs the fault type and health level. The logic for determining the fault type is as follows:

[0118] When the narrowband search peak shows a significant peak at BPFO, and the vibration kurtosis value is greater than the preset threshold, while the average value of the MFCC high-frequency component exceeds the normal range, and the extracted fundamental frequency matches BPFO, it is judged to be bearing outer ring wear or pitting.

[0119] When the narrowband search peak shows a significant peak at BPFI, and the vibration kurtosis value is greater than the preset threshold, while the MFCC mean value shifts, and the extracted fundamental frequency matches BPFI, it is judged to be bearing inner ring wear or pitting.

[0120] When the vibration kurtosis value is greater than the preset threshold, and the short-term energy is concentrated in the mid-low frequency band, the MFCC standard deviation increases significantly, but no obvious narrow band peaks appear at BPFO and BPFI, and the fundamental frequency is unstable or cannot be extracted, it is judged as wire rope derailment and swinging.

[0121] When the energy of the high-frequency sensitive subband increases significantly and the mean value of MFCC shows abnormal components in the high-frequency region, but the kurtosis value does not exceed the standard significantly, it is judged to be a crack or spalling of the anti-rope wheel body.

[0122] When the vibration kurtosis value continuously exceeds the standard and the short-term energy remains high for multiple consecutive time windows, but the narrowband search fails to match the BPFO or BPFI frequency, and the overall energy of each subband is too high, and the fundamental frequency is disordered or disappears, it is judged to be bearing jamming.

[0123] The training process of the random forest model is as follows: collect historical running data and various fault samples, extract multi-dimensional feature vectors according to the above method and label the corresponding fault types to form a training set; use the training set to train the random forest model, and adjust parameters such as the number of decision trees and maximum depth through cross-validation to optimize classification performance; after training, deploy to the cloud analysis layer.

[0124] Step 9: Tiered Early Warning

[0125] The client retrieves diagnostic results from the cloud and displays them in different levels: a green light indicates a normal health level; a yellow light indicates a need for attention and prompts for enhanced monitoring; and a red light indicates a warning and prompts for immediate repair, while also displaying the specific fault type.

[0126] This invention achieves real-time, accurate, and multi-type identification of elevator anti-corrosion sheave operation faults through coupled diagnosis of vibration and acoustic dual-mode signals, combined with an edge-cloud collaborative architecture of initial screening at the edge and in-depth analysis at the cloud. Vibration sensors detect internal mechanical impacts in the bearings, while acoustic sensors capture abnormal noises and periodic tonal characteristics of the steel wire rope. These two independent physical quantities are mutually verified through a cross-modal co-frequency verification mechanism using the fundamental frequency and fault frequency, effectively reducing false alarm and missed alarm rates. The edge-end dual-threshold mechanism significantly filters redundant data, reducing transmission and storage costs. The cloud integrates physical and acoustic features, using a random forest model to distinguish five fault types, providing fine-grained diagnosis and strong interpretability, offering maintenance personnel clear decision-making basis.

[0127] Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. Furthermore, those skilled in the art will recognize that, based on the ideas of this invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A fault diagnosis system for elevator anti-corrosion sheave operation that integrates vibration and sound information, characterized in that, It includes a signal acquisition layer, an edge gateway layer, a cloud analytics layer, and a client layer; The signal acquisition layer includes a vibration sensor installed on the elevator anti-cord pulley bearing seat or pulley support structure, and a sound sensor installed on the anti-cord pulley guard or the wire rope swing area, for synchronously acquiring the original vibration signal and the original ambient sound signal during elevator operation. The edge gateway layer is connected to the signal acquisition layer and is used to preprocess, frame, and align the original vibration signal and the original ambient sound signal in the time domain, and to filter out abnormal data segments by calculating the time domain kurtosis value of the vibration signal and the short-time energy value of the sound signal. The cloud-based analysis layer is communicatively connected to the edge gateway layer. It is used to receive the abnormal data segments, extract vibration-side features and sound-side features, fuse them, construct a multi-dimensional feature vector, and output the fault type and health level through a classification model. The constructed multi-dimensional feature vector is composed of estimated frequency shift, BPFO narrowband search peak, BPFI narrowband search peak, energy of each sensitive subband, mean value of cepstral coefficients of each order Mel frequency, standard deviation of cepstral coefficients of each order Mel frequency, mean value of fundamental frequency, and standard deviation of fundamental frequency. The client layer is connected to the cloud analysis layer and is used to obtain and display the fault type and health level.

2. The elevator anti-corrosion sheave operation fault diagnosis system integrating vibration and sound information according to claim 1, characterized in that, The preprocessing performed by the edge gateway layer includes bandpass filtering, wavelet threshold denoising, and DC component removal performed sequentially. The wavelet threshold denoising uses a soft threshold function, and the wavelet threshold is adaptively determined based on the median absolute difference of the wavelet coefficients and the signal length.

3. The elevator anti-corrosion sheave operation fault diagnosis system integrating vibration and sound information according to claim 1, characterized in that, The vibration sensor is a triaxial accelerometer configured to capture radial and axial impact pulses from the bearing.

4. The elevator anti-corrosion sheave operation fault diagnosis system integrating vibration and sound information according to claim 1, characterized in that, The cloud-based analysis layer includes a feature extraction and fusion module, which is configured to perform the following operations: Fast Fourier Transform is performed on the abnormal vibration data segment to estimate the rotational frequency, and the bearing outer ring failure frequency BPFO and bearing inner ring failure frequency BPFI are calculated based on the number of rolling elements, rolling element diameter, pitch circle diameter, contact angle and estimated rotational frequency. Search for spectral peaks at BPFO and BPFI frequencies to obtain narrowband search peaks for BPFO and BPFI. The energy of the top m sensitive subbands with the largest energy percentage is extracted by wavelet packet decomposition.

5. The elevator anti-corrosion sheave operation fault diagnosis system integrating vibration and sound information according to claim 4, characterized in that, The feature extraction and fusion module is further configured to: Extract Mel frequency cepstral coefficients from abnormal sound data segments and calculate the mean and standard deviation of Mel frequency cepstral coefficients for each order. Autocorrelation analysis or cepstral analysis is performed on the audio data segments to extract the fundamental frequency, and the standard deviation of the fundamental frequency in the abnormal data segments is calculated as a stability index.

6. The elevator anti-corrosion sheave operation fault diagnosis system integrating vibration and sound information according to claim 1, characterized in that, The cloud-based analysis layer also includes a diagnostic classification module, which has a built-in random forest classification model. The random forest classification model is configured to output the corresponding fault type and health level based on the input multidimensional feature vector.

7. A method for diagnosing elevator anti-corrosion sheave operation faults by integrating vibration and sound information, implemented based on the system described in any one of claims 1 to 6, characterized in that, include: Simultaneously collect raw vibration signals and raw ambient sound signals at the elevator anti-rope sheave; The original vibration signal and the original ambient sound signal are sequentially subjected to bandpass filtering, wavelet threshold denoising, and DC component removal to obtain the denoised signal. The denoised signal is framed and time-domain aligned according to the timestamp; Calculate the time-domain kurtosis value of the vibration signal and the short-time energy value of the sound signal. If the kurtosis value is greater than the preset kurtosis threshold and the short-time energy value is greater than the preset energy threshold, then extract the corresponding abnormal vibration data segment and abnormal sound data segment. Upload the abnormal vibration data segment and the abnormal sound data segment to the cloud; The FFT and wavelet packet decomposition algorithms are executed on the abnormal vibration data segments in the cloud to extract the narrowband search peaks and sensitive subband energy at the frequency of the frequency conversion, BPFO and BPFI. The mean and standard deviation of the Mel frequency cepstral coefficients are extracted from the abnormal sound data segments in the cloud, and the fundamental frequency and its stability index are also extracted. Vibration-side features and sound-side features are fused into a multi-dimensional feature vector, which is then input into a random forest classification model to output the fault type and health level. The client executes a tiered warning display based on the health level.

8. The elevator anti-corrosion sheave operation fault diagnosis system integrating vibration and sound information according to claim 7, characterized in that, The judgment logic of the random forest classification model includes: When a narrow-band search peak appears at BPFO, the vibration kurtosis value is greater than the preset threshold, the mean value of the high-frequency component of the Mel frequency cepstral coefficient is out of range and the fundamental frequency matches BPFO, it is determined to be bearing outer ring wear or pitting. When a narrow-band search peak appears at BPFI, the vibration kurtosis value is greater than the preset threshold, the mean value of the Mel frequency cepstral coefficients shifts and the fundamental frequency matches BPFI, it is determined to be bearing inner ring wear or pitting. When the vibration kurtosis value is greater than the preset threshold, the sound energy is concentrated in the mid-low frequency band for a short time, the standard deviation of the Mel frequency cepstral coefficient increases, there are no narrow band peaks at BPFO and BPFI and the fundamental frequency is unstable or cannot be extracted, it is determined to be a wire rope derailment and swinging. When the energy of the high-frequency sensitive subband extracted by wavelet packet decomposition increases and the mean value of the Mel frequency cepstral coefficient is abnormal in the high-frequency region but the kurtosis value does not exceed the standard, it is judged as a crack or spalling of the anti-rope wheel body. When the vibration kurtosis value continuously exceeds the standard, the short-term energy remains high in the continuous time window, the BPFO or BPFI frequency is not matched, the overall energy of each subband is too high and the fundamental frequency is disordered or disappears, it is judged as bearing jamming.

9. The elevator anti-corrosion sheave operation fault diagnosis system integrating vibration and sound information according to claim 7, characterized in that, The tiered early warning system is implemented based on the health level: when the health level is normal, the client displays a green light; when the health level is alert, the client displays a yellow light and prompts for enhanced monitoring; when the health level is warning, the client displays a red light and prompts for immediate maintenance.

Images