An elevator safety fault diagnosis method based on multi-source information fusion

Abstract

The application provides an elevator safety fault diagnosis method based on multi-source information fusion, and relates to the technical field of elevator fault diagnosis. The method first collects vibration signals of a car and a reduction gearbox and visual signals of a car traction rope during elevator operation, and extracts vibration signal features of the car and the reduction gearbox and transverse displacement signal features of the car traction rope after data preprocessing of the collected signals; then, abnormal vibration signal detection is performed according to the vibration signal features of the car and the reduction gearbox, abnormal vibration signal detection results are obtained, abnormal transverse displacement detection is performed based on the transverse displacement signal features of the car traction rope, and abnormal transverse detection results are obtained; finally, the abnormal vibration signal detection results and the abnormal transverse displacement signal detection results are fused, and whether an emergency stop fault alarm is generated for the elevator is judged based on the fusion results. The method utilizes fuzzy information fusion theory, and realizes real-time centralized monitoring of the operating state of multiple components of the elevator.

Description

Technical Field

[0001] This invention relates to the field of elevator fault diagnosis technology, and in particular to an elevator safety fault diagnosis method based on multi-source information fusion. Background Technology

[0002] With the development of urbanization, elevators are being used more and more widely to meet people's needs for vertical transportation in daily life. However, the increased use has also caused many safety accidents and hidden dangers. Elevator accidents not only cause economic losses but also seriously threaten the personal safety of passengers. Therefore, a more comprehensive and effective inspection of elevator safety is needed.

[0003] Currently, elevator fault diagnosis typically focuses on testing individual elevator components, which cannot guarantee the normal operation of multiple components simultaneously. Single-component safety assurance mainly relies on regular maintenance and annual inspections, but this approach suffers from problems such as failing to comprehensively and promptly identify elevator safety hazards and wasting human and material resources. Therefore, it is of great significance to conduct safety assessments of elevators based on online monitoring data and to achieve condition-based maintenance of elevators. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an elevator safety fault diagnosis method based on multi-source information fusion to address the shortcomings of the prior art and realize the diagnosis of elevator safety faults.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: an elevator safety fault diagnosis method based on multi-source information fusion, comprising the following steps:

[0006] Step 1: Collect vibration signals from the elevator car and gearbox, as well as visual signals from the car traction rope, during elevator operation.

[0007] The data acquisition section includes time-domain signals of elevator car and gearbox vibration and image visual signals of traction rope. The acquired time-domain signals include peak value, mean value and impulse index of the time-domain signals.

[0008] Step 2: Perform data preprocessing on the collected vibration and visual signals of the car and gearbox, and extract the vibration signal characteristics of the car and gearbox and the lateral displacement signal characteristics of the car traction rope, respectively.

[0009] The signal preprocessing section includes two parts: time-domain signal preprocessing of elevator car and gearbox vibration and image visual signal preprocessing of traction rope. Specifically:

[0010] The time-domain signals of the vibration of the elevator car and gearbox are collected and denoised using median filtering. The processed time-domain signals are then transformed into frequency-domain signals using Fourier transform. Wavelet analysis is then used to extract features from the frequency-domain signals to obtain the vibration signal features of the car and gearbox.

[0011] The collected visual signals of the car traction rope image are scanned and denoised. The current traction rope image information is calibrated using the Zhang Zhengyou calibration method. The target tracking method in computer vision is used to track the light stripe area illuminating the steel wire rope in the camera lens. Then, the lateral vibration displacement of the steel wire rope is obtained as the lateral displacement signal feature of the car traction rope.

[0012] The target tracking adopts the CSRT target tracking method, the input signal of which is the image signal of the elevator traction rope, and the output result is the image signal after virtual labeling; the image signal detection model used by the CSRT target tracking method uniformly uses CNN for classification and prediction of bounding box position;

[0013] Step 3: Detect abnormal vibration signals based on the vibration signal characteristics of the car and gearbox, obtain abnormal vibration signal detection results, and detect abnormal lateral displacement based on the lateral displacement signal characteristics of the car traction rope, and obtain abnormal lateral detection results.

[0014] An abnormal vibration signal detection model is used to detect and process the vibration signals of the elevator car and reducer, and an abnormal image state detection model is used to detect and process the image visual signals of the elevator traction rope.

[0015] The abnormal vibration signal detection model adopts the LightGBM model, whose input is the vibration signal characteristics of the car and gearbox, and whose output is the abnormal vibration signal detection result.

[0016] The abnormal image detection model adopts the R-CNN model, whose input is the image signal of the elevator car traction rope and the output is the abnormal image signal detection result. Then, the abnormal image signal is calibrated and tracked to obtain the abnormal lateral displacement signal.

[0017] Step 4: Fuse the abnormal vibration signal detection results and the abnormal lateral displacement signal detection results, and determine whether to generate an emergency stop fault alarm for the elevator based on the fusion result;

[0018] The abnormal vibration signal detection results and abnormal lateral displacement signal detection results are fused based on fuzzy information to obtain the diagnostic results, as shown in the following formula:

[0019] Y = X × R AB

[0020] in,

[0021] X = [x1 x2 x3]

[0022]

[0023] Where A represents the set of diagnosed elevator fault points, namely, three types of faults: severe vibration of the elevator car and reducer, deformation of the car traction rope exceeding the safety value, etc.; B represents the set of elevator operating states, namely, three types of elevator operating state signals: vibration signals of the car and reducer, and lateral displacement signals of the traction rope; the relationship matrix R between A and B is... AB element μ ij Let X be the probability of inferring the fault type j from the operating state i of different elevator components, i = 1, 2, 3, j = 1, 2, 3, X be the set of confidence values ​​for judging the operating state of different elevator components, x1, x2, x3 be the confidence values ​​assigned to the operating state of different elevator components based on experience or fault database, and Y, obtained by fuzzy transformation, be the probability of each decision after fusion.

[0024] When making inference decisions about faults, a rule-based approach is adopted, with the following basic principles:

[0025] (1) The decision objective should have the largest membership value;

[0026] (2) The membership value of the decision objective must be greater than the set threshold;

[0027] (3) The difference between the membership values ​​of the decision objective and other objectives must be greater than a certain threshold;

[0028] Finally, the possible sets of each fault inference decision are selected according to certain weights to obtain the optimal fault inference result.

[0029] The beneficial effects of adopting the above technical solution are as follows: The elevator safety fault diagnosis method based on multi-source information fusion provided by the present invention, which uses fuzzy information fusion theory to realize real-time centralized monitoring of the operating status of multiple elevator components, targeting the operating status of elevator car, gearbox and traction rope, and assigns different weights to the fault decisions of different components, thereby improving the speed and accuracy of elevator fault diagnosis. Attached Figure Description

[0030] Figure 1 A flowchart illustrating an elevator safety fault diagnosis method based on multi-source information fusion, provided as an embodiment of the present invention;

[0031] Figure 2 A schematic diagram of the process of the elevator safety fault diagnosis method based on multi-source information fusion provided in the embodiments of the present invention.

[0032] Figure 3This is a flowchart illustrating the fault diagnosis results obtained based on fuzzy information fusion, as provided in an embodiment of the present invention. Detailed Implementation

[0033] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0034] In this embodiment, an elevator safety fault diagnosis method based on multi-source information fusion is described, such as... Figure 1 , 2 As shown, it includes the following steps:

[0035] Step 1: Collect vibration signals from the elevator car and gearbox, as well as visual signals from the car traction rope, during elevator operation.

[0036] The data acquisition section includes time-domain signals of elevator car and gearbox vibration and image visual signals of traction rope. The acquired time-domain signals include peak value, mean value and impulse index of the time-domain signals.

[0037] In this embodiment, the elevator vibration signal monitoring points are arranged as follows: the elevator vibration status monitoring points are arranged on the car and the gearbox respectively. The sensor nodes inside the car are mainly arranged at the center of the three-way wall of the car. The elevator vibration signal is collected by installing independent acceleration sensors on the outside of the elevator mechanical components.

[0038] Image status monitoring points for the elevator car's traction rope are installed on the elevator shaft wall. A line-structured light-based detection method is used, with an imaging system comprising a line light source and an industrial CCD camera. The line light source and camera have relatively fixed spatial positions, and the light plane of the line light source is perpendicular to the axis of the traction rope. When the laser plane of the linear structured vision sensor is projected onto the surface of the workpiece to be measured, the laser plane produces deformed light streaks due to changes in the workpiece's surface dimensions. The camera captures the image of these deformed light streaks. This method uses independent sensor acquisition and has no impact on the existing mechanical structure and control circuitry of the elevator, nor does it affect the normal operation of the elevator.

[0039] Step 2: Perform data preprocessing on the collected vibration and visual signals of the car and gearbox, and extract the vibration signal characteristics of the car and gearbox and the lateral displacement signal characteristics of the car traction rope, respectively.

[0040] The signal preprocessing section includes two parts: time-domain signal preprocessing of elevator car and gearbox vibration and image visual signal preprocessing of traction rope. Specifically:

[0041] The time-domain signals of the vibration of the elevator car and gearbox are collected and denoised using median filtering. The processed time-domain signals are then transformed into frequency-domain signals using Fourier transform. Wavelet analysis is then used to extract features from the frequency-domain signals to obtain the vibration signal features of the car and gearbox.

[0042] The collected visual signals of the car traction rope image are scanned and denoised. The current traction rope image information is calibrated using the Zhang Zhengyou calibration method. The target tracking method in computer vision is used to track the light stripe area illuminating the steel wire rope in the camera lens. Then, the lateral vibration displacement of the steel wire rope is obtained as the lateral displacement signal feature of the car traction rope.

[0043] The target tracking employs the CSRT target tracking method. Its input signal is the image (video) signal of the elevator traction rope, and the output is the image (video) signal after virtual labeling. The image signal detection model used in the CSRT target tracking method uniformly uses CNN for classification and bounding box position prediction. CNN is a seven-layer convolutional neural network, a type of feedforward neural network algorithm that includes convolutional computation and has a deep structure. Its first layer is the input layer; the second and third layers are convolutional layers used for image feature extraction, converting the image into a fixed-dimensional feature vector; the fourth, fifth, and sixth layers are pooling layers used for feature compression and extraction of key features; the seventh layer is a fully connected layer that classifies the original image using the extracted feature parameters; and the eighth layer is the output layer.

[0044] Step 3: Detect abnormal vibration signals based on the vibration signal characteristics of the car and gearbox, obtain abnormal vibration signal detection results, and detect abnormal lateral displacement based on the lateral displacement signal characteristics of the car traction rope, and obtain abnormal lateral detection results.

[0045] An abnormal vibration signal detection model is used to detect and process the vibration signals of the elevator car and reducer, and an abnormal image state detection model is used to detect and process the image visual signals of the elevator traction rope.

[0046] The abnormal vibration signal detection model adopts the LightGBM model, whose input is the vibration signal characteristics of the car and gearbox, and whose output is the abnormal vibration signal detection result.

[0047] The abnormal image detection model adopts the R-CNN model, whose input is the image signal of the elevator car traction rope and the output is the abnormal image signal detection result. Then, the abnormal image signal is calibrated and tracked to obtain the abnormal lateral displacement signal.

[0048] Step 4: Fuse the abnormal vibration signal detection results and the abnormal lateral displacement signal detection results, and determine whether to generate an emergency stop fault alarm for the elevator based on the fusion result;

[0049] In this embodiment, fault diagnosis based on fuzzy information fusion is performed on the abnormal vibration signal detection results and the abnormal lateral displacement signal detection results. Figure 3 As shown, specifically:

[0050] First, elevator operating status data is collected and fuzzified to establish a relationship matrix between operating status and fault type. The weight vectors of each sensor are then obtained and multiplied with the relationship matrix to obtain the diagnostic membership result matrix. Finally, based on the fault determination principle, the probability of each fused fault decision is obtained, as shown in the following formula:

[0051] Y = X × R AB

[0052] in,

[0053] X = [x1 x2 x3]

[0054]

[0055] Where A represents the set of diagnosed elevator fault points, namely, three types of faults: severe vibration of the elevator car and reducer, deformation of the car traction rope exceeding the safety value, etc.; B represents the set of elevator operating states, namely, three types of elevator operating state signals: vibration signals of the car and reducer, and lateral displacement signals of the traction rope; the relationship matrix R between A and B is... AB element μ ij Let X be the probability of inferring the fault type j from the operating state i of different elevator components, i = 1, 2, 3, j = 1, 2, 3, X be the set of confidence values ​​for judging the operating state of different elevator components, x1, x2, x3 be the confidence values ​​assigned to the operating state of different elevator components based on experience or fault database, and Y, obtained by fuzzy transformation, be the probability of each decision after fusion.

[0056] When making inference decisions about faults, a rule-based approach is adopted, with the following basic principles:

[0057] (1) The decision objective should have the largest membership value;

[0058] (2) The membership value of the decision objective must be greater than the set threshold;

[0059] (3) The difference between the membership values ​​of the decision objective and other objectives must be greater than a certain threshold;

[0060] Finally, the possible sets of each fault inference decision are selected according to certain weights to obtain the optimal fault inference result.

[0061] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the claims of the present invention.

Claims

1. A method for diagnosing elevator safety faults based on multi-source information fusion, characterized in that: Includes the following steps: Step 1: Collect vibration signals from the elevator car and gearbox, as well as visual signals from the car traction rope, during elevator operation. Step 2: Perform data preprocessing on the collected vibration and visual signals of the car and gearbox, and extract the vibration signal characteristics of the car and gearbox and the lateral displacement signal characteristics of the car traction rope, respectively. The data preprocessing includes two parts: time-domain signal preprocessing of the elevator car and gearbox vibration and image visual signal preprocessing of the traction rope. Specifically: The time-domain signals of the vibration of the elevator car and gearbox are collected and denoised using median filtering. The processed time-domain signals are then transformed into frequency-domain signals using Fourier transform. Wavelet analysis is then used to extract features from the frequency-domain signals to obtain the vibration signal features of the car and gearbox. The collected visual signals of the car traction rope image are scanned and denoised. The current traction rope image information is calibrated using the Zhang Zhengyou calibration method. The target tracking method in computer vision is used to track the light stripe area illuminating the steel wire rope in the camera lens. Then, the lateral vibration displacement of the steel wire rope is obtained as the lateral displacement signal feature of the car traction rope. Step 3: Detect abnormal vibration signals based on the vibration signal characteristics of the car and gearbox, obtain abnormal vibration signal detection results, and detect abnormal lateral displacement based on the lateral displacement signal characteristics of the car traction rope, and obtain abnormal lateral detection results. An abnormal vibration signal detection model is used to detect and process the vibration signals of the elevator car and reducer, and an abnormal image state detection model is used to detect and process the image visual signals of the elevator traction rope. The abnormal vibration signal detection model adopts the LightGBM model, whose input is the vibration signal characteristics of the car and gearbox, and whose output is the abnormal vibration signal detection result. The abnormal image detection model adopts the R-CNN model, whose input is the image signal of the elevator car traction rope and the output is the abnormal image signal detection result. Then, the abnormal image signal is calibrated and tracked to obtain the abnormal lateral displacement signal. Step 4: Fuse the abnormal vibration signal detection results and the abnormal lateral displacement signal detection results, and determine whether to generate an emergency stop fault alarm for the elevator based on the fusion result; Fault diagnosis based on fuzzy information fusion is performed on the abnormal vibration signal detection results and abnormal lateral displacement signal detection results. Specifically, the elevator operating status data is first collected and fuzzified to establish a relationship matrix between operating status and fault type. The weight vectors of each sensor are then obtained and multiplied with the relationship matrix to obtain the diagnostic membership result matrix. Finally, the probability of each fault decision after fusion is obtained according to the fault judgment principle, as shown in the following formula: Y = X × RAB in, in, A The set of elevator fault points to be diagnosed includes three types of faults: severe vibration of the elevator car and reducer, deformation of the car traction rope exceeding the safe value, etc. B It is a set of elevator operating status signals, namely, the vibration signals of the car and gearbox, and the lateral displacement signal of the traction rope. A and B relation matrix RAB elements in μij The operating status of different elevator components i Inference decision failure type j The possibility, , , X This is a set of reliability values ​​for judging the operating status of different components of each elevator. These are confidence values ​​assigned to the operating states of different elevator components based on experience or fault databases, obtained through fuzzy transformation. Y To determine the possibilities for each decision after integration; Finally, the possible sets of each fault inference decision are selected according to certain weights to obtain the optimal fault inference result.

2. The elevator safety fault diagnosis method based on multi-source information fusion according to claim 1, characterized in that: Step 1 involves acquiring time-domain signals of the elevator car and gearbox vibrations, as well as visual images of the traction rope. The acquired time-domain signals include the peak value, mean value, and pulse index of the time-domain signals.

3. The elevator safety fault diagnosis method based on multi-source information fusion according to claim 2, characterized in that: The target tracking adopts the CSRT target tracking method, the input signal of which is the image signal of the elevator traction rope, and the output result is the image signal after virtual labeling; the image signal detection model used in the CSRT target tracking method uniformly uses CNN for classification and prediction of bounding box position.

4. The elevator safety fault diagnosis method based on multi-source information fusion according to claim 3, characterized in that: The operating status of different elevator components i Inference decision failure type j When there is a possibility of [something], a rule-based approach is adopted, with the following basic principles: (1) The decision objective to be determined should have the highest membership value; (2) The membership value of the decision objective must be greater than the set threshold; (3) The difference between the membership values ​​of the decision target and other targets must be greater than a certain threshold.

Images