Method and device for evaluating service life of composite steel belt of elevator

Abstract

The invention relates to the technical field of equipment detection, in particular to an elevator composite steel belt service life evaluation method and device, which comprises a steel belt, a steel belt traction wheel and a steel belt traction machine, and is characterized in that the periphery of the steel belt is provided with a steel belt front industrial camera, a steel belt right industrial camera, a steel belt left industrial camera and a steel belt back industrial camera respectively; a vibration sensor is arranged on the top of the steel belt traction machine. The method comprises the following steps: acquiring an image of the surface of the steel strip through an industrial camera, preprocessing and extracting features of the image, and judging whether the surface defects and the thickness of the steel strip meet scrap technical conditions or not; judging whether the bending frequency and the service time of the steel belt reach the allowable simple bending frequency and declaration age limit or not through a vibration sensor and a timer; by recording the number of bending times of the steel belt, the actually measured thickness of the steel belt and the load bearing of the elevator, the BP neural network algorithm is used for predicting the time when the steel belt reaches service data, and the remaining service life of the steel belt is predicted.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment detection, and specifically to a method and device for evaluating the life of an elevator composite steel belt. Background Technique

[0002] As a traction tool for a new generation of elevators, the composite steel belt is usually composed of a polyurethane material covering multiple flat steel wire rope cores, and the surface is generally flat or toothed. Compared with the traditional steel wire rope traction method, the polyurethane surface layer directly contacts the traction wheel, which has the advantages of quietness, environmental protection, and energy saving, while maximizing the use of the hoistway space. In addition, the contact area between the steel belt and the traction wheel is larger, providing sufficient friction, avoiding the direct contact between the steel wire rope and the traction wheel, not only protecting the internal steel wire rope core but also extending the service life of the traction wheel, thus greatly improving the overall durability of the elevator. Given the advantages of the composite steel belt, steel belt elevators have been widely used. However, with the long-term service of the steel belt, the traction steel belt may be damaged and worn, bringing potential safety hazards to the safe operation of the elevator. Therefore, the monitoring of the steel belt scrapping and the prediction of the remaining service life have become the key to ensuring the safe and reliable operation of the elevator.

[0003] According to the national standard GB / T 39172-2020 "Non-steel wire rope suspension devices for elevators", when the steel belt shows any of the following conditions, it is considered to reach the scrapping technical conditions: 1) The coating layer of the steel belt is deformed, such as bulging, indentation, crease, depression, etc.; 2) The internal steel wire rope core is exposed due to cracks or wear in the coating layer of the steel belt; 3) The steel wire rope core pierces out on the surface of the coating layer of the steel belt; 4) The measured thickness of the steel belt is reduced to the specified value of the manufacturer relative to the nominal thickness; 5) The number of bending times and the service time of the steel belt reach the allowable simple bending times or the declared years.

[0004] Currently, the judgment of the scrapping of elevator composite steel belts mainly focuses on visual surface defect detection and non-destructive detection based on electromagnetic principles. Patent ZL201921802836.8 invented a device for detecting the state of the internal steel wire rope of an elevator traction steel belt through an electromagnetic detection component. Patent ZL202210672496.1 invented an intelligent elevator steel belt surface defect detection device for detecting the surface defects of elevator steel belts. Patent ZL201910271604.2 invented an online health prediction method for elevator composite steel belts. According to the operating parameters of the elevator, this patent uses a fretting wear model to predict the damage of the composite steel belt and the resistance change of the steel wire rope embedded in the composite steel belt, and then outputs the health prediction result of the steel belt.

[0005] At present, the scrapping determination of steel belts mainly adopts machine vision and electromagnetic methods, only considering surface defects and internal wire rope breakage, lacking multi-dimensional scrapping determination of steel belts; and for predicting the remaining service life of elevator steel belts, the scrapping caused by surface wear of the steel belt coating has not been considered. Summary of the Invention

[0006] The purpose of the present invention is to provide a method and device for evaluating the life of an elevator composite steel belt. By visually inspecting the surface of the steel belt, identifying surface defects of the steel belt and measuring the thickness of the steel belt, and by measuring the number of bends and service time of the steel belt, it is evaluated whether the steel belt meets the scrapping technical conditions, and the remaining service life is predicted according to the measured number of bends of the steel belt, the measured thickness of the steel belt, and the load of the elevator, so as to solve the problems of difficult real-time determination of the scrapping of elevator composite steel belts and prediction of the remaining life.

[0007] To achieve the above purpose, the present invention provides the following technical solution: An elevator composite steel belt life evaluation device includes a steel belt, a steel belt traction wheel, and a steel belt traction machine. Industrial cameras are respectively arranged on the front, right side, left side, and back of the steel belt, and a vibration sensor is arranged on the top of the steel belt traction machine;

[0008] The vibration sensor judges the start-up times of the steel belt traction machine by monitoring the change of the motor vibration signal, and then infers the number of bends of the steel belt.

[0009] Preferably, the elevator composite steel belt life evaluation device further includes a timer installed in the elevator control cabinet and a load sensor installed in the elevator. The timer records the service time of the steel belt and gives an early warning when approaching the predetermined life of the steel belt, and the load sensor is used to measure the load of the elevator.

[0010] The present invention also provides a method for evaluating the life of an elevator composite steel belt. The method includes the following steps:

[0011] S1: Obtain the surface condition of the steel belt through the industrial cameras installed around the steel belt, and perform denoising, image conversion, and feature extraction on the image;

[0012] S2: According to the extracted geometric shape features, texture features, and edge features, use a classifier to identify surface defects and judge whether the scrapping technical conditions are met;

[0013] S3: The extracted edge features are also used to calculate the measured thickness of the steel belt, and it is judged whether to scrap according to whether the measured thickness is reduced to the specified value provided by the manufacturer relative to the nominal thickness;

[0014] S4: By recording the service time and bending times of the steel strip, determine whether it is scrapped based on whether the allowable simple bending times and declared life are reached;

[0015] S5: Based on the measured bending times of the steel belt, the actual thickness of the steel belt, and the load-bearing capacity of the elevator, the BP neural network algorithm is used to predict the time when the steel belt reaches the service data and its remaining service life.

[0016] Preferably, in step S1, for image denoising, the mean filtering principle is adopted, and the image is denoised by calculating the average value in the neighborhood of the pixel point and using the average value of the grayscale values ​​of all pixels to replace the original pixels. For image conversion, the image is converted using the weighted average method, and the color image is converted into an image with only grayscale levels, thereby enhancing the features in the image and making the defects more obvious, which is convenient for subsequent feature extraction and analysis.

[0017] Preferably, in step S2, the geometric shape feature uses two geometric parameters, area and perimeter, to describe the characteristics of the surface defects of the steel strip. The area parameter is calculated by traversing each pixel point in the defect area and adding their values, which is the total number of pixels in the defect area. The perimeter parameter is obtained by calculating the total number of pixels on the boundary of the defect area.

[0018] Preferably, in step S2, texture features are used to describe complex patterns on the surface of an object, and a local binary pattern is used to capture the texture.

[0019] Preferably, in step S2, the edge feature provides information about the boundary of the object, which is convenient for identifying the shape and outline of the object. The Canny edge detection algorithm is used to select a suitable detection window size, and the gradients in the horizontal and vertical directions are calculated respectively by the Sobel operator. Then, the amplitude and direction of the gradient are calculated by the gradient, and the gradient amplitude is non-maximum suppressed to eliminate false edge information. At the same time, the edges are classified using double thresholds.

[0020] Preferably, in step S2, the classifier uses a support vector machine, and the SVM automatically detects the type and location of the surface defects of the steel strip according to the input defect feature vector through training and learning, and determines whether the defects reach the scrap technical conditions.

[0021] Preferably, in step S3, the actual measured thickness of the steel strip is calculated by selecting reference points corresponding to the same height of the steel strip on both side edges of the contour as much as possible, calculating the Euclidean distance between the reference points, and converting the actual measured thickness of the steel strip by the proportional factor between the distance in the image and the actual thickness. Since the steel strip may deviate during the operation of the elevator, the point with the minimum thickness is selected as the actual measured thickness of the steel strip, and whether the steel strip is scrapped is determined based on whether the actual measured thickness is reduced to the specified value provided by the manufacturer relative to the nominal thickness.

[0022] Preferably, in step S5, each time the elevator motor starts, the number of bends, thickness, and elevator load-bearing of the steel belt are recorded once, and the recorded data is used as the input of the BP neural network to predict the remaining service life of the steel belt.

[0023] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0024] 1. The present invention judges whether the steel belt reaches the scrap technical conditions through various methods, and at the same time has a certain real-time monitoring and prediction ability. The surface defects and thickness changes of the steel belt are monitored through machine vision, and the use time and number of bends of the steel belt are recorded at the same time to judge whether the steel belt reaches the scrap technical conditions; the BP neural network is used to predict the remaining service life of the steel belt.

[0025] 2. The present invention can realize real-time monitoring of the scrap technical conditions such as the surface defects and thickness of the steel belt, as well as prediction of the remaining service life of the steel belt, helping maintenance personnel to judge the scrap of the steel belt and predict its life, and ensuring the safety and reliability of the operation of the steel belt elevator. Description of the Drawings

[0026] Figure 1 It is a schematic diagram of the overall structure of the elevator composite steel belt scrap determination and life prediction device of the present invention;

[0027] Figure 2 It is a steel belt scrap determination process based on visual inspection provided by the present invention;

[0028] Figure 3 It is a steel belt scrap determination process based on visual measurement provided by the present invention;

[0029] Figure 4 It is a scrap determination process based on the service time of the steel belt provided by the present invention;

[0030] Figure 5 It is a scrap determination process based on the number of bends of the steel belt provided by the present invention;

[0031] Figure 6 It is a prediction process for the remaining service life of the steel belt based on the BP neural network provided by the present invention;

[0032] Figure 7 It is the network structure of the BP neural network.

[0033] In the figure: 1, industrial camera on the front side of the steel belt; 2, industrial camera on the right side of the steel belt; 3, industrial camera on the left side of the steel belt; 4, industrial camera on the back side of the steel belt; 5, steel belt; 6, steel belt traction wheel; 7, steel belt traction machine; 8, vibration sensor. Detailed Embodiment

[0034] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0035] Embodiment 1:

[0036] An elevator composite steel belt life evaluation device includes a steel belt 5, a steel belt traction wheel 6, and a steel belt traction machine 7. An industrial camera 1 on the front side of the steel belt, an industrial camera 2 on the right side of the steel belt, an industrial camera 3 on the left side of the steel belt, and an industrial camera 4 on the back side of the steel belt are respectively arranged around the steel belt 5. A vibration sensor 8 is arranged on the top of the steel belt traction machine 7;

[0037] The vibration sensor 8 judges the starting times of the steel belt traction machine 7 by monitoring the change of the motor vibration signal, and then infers the bending times of the steel belt 5.

[0038] In this embodiment, the elevator composite steel belt life evaluation device further includes a timer installed in the elevator control cabinet and a load sensor installed in the elevator. It is characterized in that: the timer records the service time of the steel belt 5 and gives an early warning when approaching the predetermined life of the steel belt 5, and the load sensor is used to measure the load of the elevator.

[0039] The present invention also provides an elevator composite steel belt life evaluation method, and the method includes the following steps:

[0040] S1: Obtain the surface condition of the steel belt through the industrial cameras installed around the steel belt, and perform denoising, image conversion, and feature extraction on the image;

[0041] S2: According to the extracted geometric shape features, texture features, and edge features, use a classifier to identify surface defects and judge whether the scrapping technical conditions are met;

[0042] S3: The extracted edge features are simultaneously used to calculate the measured thickness of the steel belt, and it is judged whether to scrap according to whether the measured thickness is reduced to the specified value provided by the manufacturer relative to the nominal thickness;

[0043] S4: By recording the service time and bending times of the steel belt, it is judged whether to scrap according to whether the allowable simple bending times and the declared years are reached;

[0044] S5: According to the measured bending times of the steel belt, the measured thickness of the steel belt, and the load of the elevator, use the BP neural network algorithm to predict the time when the steel belt reaches the service data and predict its remaining service life.

[0045] In this embodiment, in step S1, for image denoising, the mean filtering principle is adopted. By calculating the average value within the neighborhood of the pixel points and using the average value of all pixel gray values to replace the original pixels, the image is denoised. For image conversion, the weighted average method is used to perform conversion processing on the image, converting the color image into an image with only gray levels, enhancing the features in the image, making the defects more obvious, and facilitating subsequent feature extraction and analysis. For feature extraction, it can help the machine learning model identify the defects in the image and distinguish them from the normal parts. At the same time, by extracting key features, the recognition accuracy of the model is improved.

[0046] In this embodiment, in step S2, the geometric shape features use two geometric parameters, area and perimeter, to describe the features of the steel strip surface defects. The calculation of the area parameter is obtained by traversing each pixel point within the defect area and adding their values together, which is the total number of pixel points within the defect area. The perimeter parameter is obtained by calculating the total number of pixel points on the boundary of the defect area.

[0047] In this embodiment, in step S2, the texture features are used to describe the complex patterns on the surface of the object. Extracting texture features is helpful for identifying defects with specific texture patterns (such as cracks, scratches, corrosion, etc.). The local binary pattern (LBP) is used to capture the texture. LBP is an effective feature extraction method for texture classification and image analysis. The basic idea is to encode the local area of each pixel and its surrounding neighborhood to capture the texture features of the image.

[0048] In this embodiment, in step S2, the edge features provide information about the object boundary, facilitating the identification of the shape and contour of the object. The Canny edge detection algorithm is used. The appropriate detection window size is selected. The gradients in the horizontal and vertical directions are calculated respectively through the Sobel operator, and then the magnitude and direction of the gradient are calculated through the gradient. Non-maximum suppression is performed on the gradient magnitude to eliminate false edge information. At the same time, double thresholds are used to classify the edges.

[0049] In this embodiment, in step S2, the classifier uses a support vector machine (SVM). Through training and learning, the SVM automatically detects the type and location of the steel strip surface defects according to the input defect feature vectors and determines whether the defects meet the scrap technical conditions.

[0050] In this embodiment, in step S3, the measured thickness of the steel strip is calculated by selecting reference points corresponding to the same height of the steel strip on both edges of the contour as much as possible, calculating the Euclidean distance between the reference points, and obtaining the measured thickness of the steel strip through conversion using the scale factor between the distance in the image and the actual thickness. Since the steel strip will shift during the operation of the elevator, the minimum thickness is selected as the measured thickness of the steel strip, and it is determined whether the steel strip is scrapped based on whether the measured thickness is reduced to the specified value provided by the manufacturer relative to the nominal thickness.

[0051] In this embodiment, in step S5, each time the elevator motor starts, the number of bends, thickness, and elevator load of the steel strip are recorded once, and the recorded data is used as the input of the BP neural network to predict the remaining service life of the steel strip.

[0052] Embodiment 2:

[0053] The method for determining the scrapping and predicting the life of the elevator composite steel strip based on machine vision proposed by the present invention mainly includes:

[0054] (1) Judging the scrapping of the steel strip based on visual detection

[0055] It is difficult to directly observe the surface condition of the steel strip during its service. Through industrial cameras fixed on the front and back surfaces facing the steel strip, the front industrial camera 1 of the steel strip and the back industrial camera 4 of the steel strip collect images of the steel strip surface, and then judge whether the scrapping technical conditions are met. However, during the acquisition process, it will be affected by various uncertain factors, making the image contain noise, which will affect the subsequent detection. Therefore, it is necessary to preprocess the acquired image, extract important features, and make the judgment more accurate. The judgment process is as follows Figure 2 .

[0056] 1.1 Image denoising

[0057] Noise will affect the visual effect of the image and the performance of subsequent detection tasks. The invention adopts the mean filtering principle. By calculating the average value within the neighborhood of the pixel point and using the average value of all pixel gray values to replace the original pixel, the image is denoised. The formula is as follows:

[0058]

[0059] In formula (1), g(i,j) is the pixel point after mean filtering of the image, M is the number of pixel points within the neighborhood, and f(m,n) is the gray value of the pixel point within the neighborhood.

[0060] 1.2 Image conversion

[0061] To better extract features and enhance the performance of the model in the follow-up, the invention uses the weighted average method to perform conversion processing on the image, converting the color image into an image with only gray levels. The formula is as follows:

[0062] f(i,j) = 0.299R + 0.587G + 0.114B (2)

[0063] In formula (2), f(i,j) is the gray value of the pixel point (i,j) of the defective color image, and R, G, and B are the pixel values of the red, green, and blue channels of the pixel point respectively.

[0064] 1.3 Feature Extraction

[0065] Feature extraction can help the machine learning model identify defects in the image and distinguish them from the normal parts. At the same time, by extracting key features, the recognition accuracy of the model can be improved. The present invention mainly extracts geometric shape features, texture features, and edge features.

[0066] 1) Geometric Shape Features

[0067] Shape features are one of the most direct features of the defective area and can effectively describe the differences between different defective areas. The present invention uses two geometric parameters, area and perimeter, to describe the features of the steel strip surface defects. The calculation of the area parameter is obtained by traversing each pixel point in the defective area and adding their values together, which is the total number of pixel points in the defective area. The perimeter parameter is obtained by calculating the total number of pixel points on the boundary of the defective area. The formula is as follows:

[0068]

[0069] L = 2(N + M) (4)

[0070] In formulas (3) and (4), I(x,y) represents the pixel value in the defective area, and M and N are the numbers of pixel points with even and odd image chain code values respectively, representing the vertical boundary and the horizontal boundary respectively.

[0071] 2) Texture Features

[0072] Texture features are used to describe the complex patterns on the surface of an object. Extracting texture features is helpful for identifying defects with specific texture patterns (such as cracks, scratches, corrosion, etc.). The invention uses Local Binary Patterns (LBP) to extract texture features. LBP is an effective feature extraction method for texture classification and image analysis. The basic idea is to encode the local area of each pixel and its surrounding neighborhood to capture the texture features of the image. The formula is as follows:

[0073]

[0074] In formula (5), (xc, yc) is the position of the central pixel, P is the number of pixels in the neighborhood, and i p is the gray value of the p-th pixel in the neighborhood, and i c is the gray value of the central pixel, and s is the sign function, defined as:

[0075]

[0076] Formula (6) compares each pixel in the neighborhood with the central pixel and converts the comparison result (1 or 0) into a binary number, finally obtaining the LBP value of the central pixel. In this way, LBP provides a feature pattern for measuring the neighborhood relationship between pixels and can effectively extract the local features of the image.

[0077] 3) Edge features

[0078] Edge features provide information about the boundaries of objects, facilitating the recognition of the shape and contour of objects. The invention uses the Canny edge detection algorithm, selects an appropriate detection window size, calculates the gradients in the horizontal and vertical directions respectively through the Sobel operator, and then calculates the magnitude and direction of the gradient through these gradients. The formulas are as follows:

[0079]

[0080] In formulas (7) and (8), M and θ represent the gradient magnitude and gradient direction respectively, and G X and G y are the gradients in the horizontal and vertical directions respectively.

[0081] Non-maximum suppression is performed on the gradient magnitude to eliminate false edge information. This process retains the maximum value of the gradient intensity at each pixel point and filters out other values, making the edges clearer. At the same time, double thresholds are used to classify the edges, and strong and weak edges are connected through hysteresis processing. The high threshold is used to find the definite edges, and then the low threshold is used to find the potential edges. If the potential edges are connected to the strong edges, they are considered actual edges. Finally, the boundary is tracked through the hysteresis technique. If a weak edge at a certain pixel position is connected to a strong boundary, it is considered a boundary, and other weak edges are deleted to ensure the continuity of the edges.

[0082] According to the extracted defect features, a classifier is used to detect and classify the defects on the steel strip surface. The classifier automatically detects the type and position of the defects on the steel strip surface according to the input feature vector through training and learning.

[0083] Support Vector Machine (SVM) is a supervised learning algorithm mainly used for classification problems. The core idea is to find a hyperplane that can maximize the separation of data points of different classes. For non-linearly separable data, SVM maps the data to a high-dimensional space by using a kernel function, making the data linearly separable in the high-dimensional space. According to the extracted features, the support vector machine is used for classification and recognition to judge the type of defects and whether the scrapping technical conditions are met.

[0084] (2) Judgment of Steel Belt Scrap Based on Visual Measurement

[0085] It is very difficult to directly measure the thickness of the steel belt during its service. In this invention, the measured thickness of the steel belt is obtained through industrial cameras 2 and 3 installed on the left and right sides of the steel belt. It is judged whether to scrap according to whether the measured thickness is reduced to the specified value provided by the manufacturer relative to the nominal thickness. During the measurement process, the obtained steel belt image is preprocessed and edge detected to obtain the contour of the steel belt. By calculating the vertical distance of each point on the contour, the actual thickness of the steel belt is converted as the measured thickness of the steel belt according to the proportional relationship between the distance in the image and the actual thickness. Since the steel belt will shift during the operation of the elevator, the minimum thickness is selected as the measured thickness of the steel belt.

[0086] Images of both sides of the steel belt are obtained using industrial cameras. The collected image data is preprocessed. Mean filtering and grayscale conversion are used to reduce image noise and reduce the level of detail, so as to improve the quality and processing efficiency of the data, reduce measurement errors, and ensure the accuracy of subsequent processing. Extract the edges or landmark points on both sides of the steel belt from the preprocessed data, which is achieved through image processing technology. The Canny edge detection algorithm is used to identify the edge points of the steel belt, such as Figure 3 .

[0087] 2.1 Image Denoising

[0088] Noise will affect the visual effect of the image and the performance of subsequent detection tasks. The invention adopts the principle of mean filtering. By calculating the average value within the neighborhood of the pixel point and using the average value of all pixel gray values to replace the original pixel, the image is denoised. The formula is as follows:

[0089]

[0090] In formula (1), g(i,j) is the pixel point after mean filtering of the image, M is the number of pixel points within the neighborhood, and f(m,n) is the gray value of the pixel points within the neighborhood.

[0091] 2.2 Image Conversion

[0092] To better extract features and enhance the performance of the model in the future, the weighted average method is used to transform the image, converting the color image into an image with only gray levels. The formula is as follows:

[0093] f(i,j) = 0.299R + 0.587G + 0.114B (2)

[0094] In Equation (2), f(i,j) is the gray value of the pixel point (i,j) in the defective color image, and R, G, and B are the pixel values of the red, green, and blue channels of the pixel point, respectively.

[0095] 2.3 Edge Detection

[0096] Edge features provide information about the boundaries of objects, facilitating the identification of object shapes and contours. The invention uses the Canny edge detection algorithm, selects an appropriate detection window size, calculates the gradients in the horizontal and vertical directions respectively through the Sobel operator, and then calculates the magnitude and direction of the gradient based on these gradients. The formula is as follows:

[0097]

[0098] In Equations (7) and (8), M and θ represent the gradient magnitude and gradient direction respectively, and G X , G y are the gradients in the horizontal and vertical directions respectively.

[0099] Non-maximum suppression is performed on the gradient magnitude to eliminate false edge information. This process retains the maximum value of the gradient intensity at each pixel point and filters out other values, making the edges clearer. At the same time, double thresholds are used to classify the edges, and the strong edges and weak edges are connected through hysteresis processing. The high threshold is used to find the definite edges, and then the low threshold is used to find the potential edges. If the potential edges are connected to the strong edges, they are considered actual edges. Finally, the boundary is tracked through the hysteresis technique. If a weak edge at a certain pixel position is connected to a strong boundary, it is considered a boundary, and other weak edges are deleted to ensure the continuity of the edges.

[0100] 2.4 Thickness Calculation

[0101] Reference points are selected on the two side edges of the contour. These points should correspond to the same height of the steel strip as much as possible to ensure the accuracy of the measurement. The shortest straight-line distance between these reference points is calculated by calculating the Euclidean distance between two points. The formula is as follows:

[0102]

[0103] In Equation (9), A(x 1 ,y 1 ), B(x 2, y 2 ) are two reference points, and d(A, B) is the Euclidean distance.

[0104] In the case where the actual thickness of the steel strip is known, several points are measured actually to establish the proportional relationship between the distance in the image and the actual thickness. According to the actual measurement results, the proportional factor between the distance in the image and the actual thickness is calculated. Using the established proportional relationship, the shortest distance calculated in the image is converted into the actual thickness of the steel strip.

[0105] Finally, the average value of all the calculated thickness values is taken as the measured thickness of the steel strip. At the same time, the outlier is removed using the statistical method (3σ principle) to improve the reliability of the measurement result. It is judged whether the steel strip is scrapped according to whether the measured thickness is reduced to the specified value provided by the manufacturer relative to the nominal thickness.

[0106] (3) Scrapping judgment based on the service time and the number of bends of the steel strip

[0107] During the service process of the steel strip, due to repeated bending and stretching, the internal structure will gradually accumulate fatigue damage. This kind of damage may not be easily detected in the initial stage, but with the increase of the service time and the number of bends, the damage will gradually accumulate, and finally may lead to a significant decrease in the strength and durability of the steel strip. When the performance no longer meets the requirements of relevant standards, the steel strip should be scrapped. If the steel strip with a long service time and a large number of bends is continued to be used, there are certain potential safety hazards.

[0108] 3.1 Scrapping judgment based on the service time of the steel strip

[0109] In the present invention, a timer is installed on the steel strip elevator to detect the installation time of the steel strip, which can help the maintenance team to understand the usage situation of the steel strip in time, so as to better arrange regular inspections and replacements and ensure the safe operation of the elevator.

[0110] By installing a timer inside the control cabinet of the steel strip elevator, in addition to the basic timing function, the timer also supports functions such as remote monitoring, data recording, and alarm, and can remind the maintenance personnel when the steel strip is approaching or reaching its predetermined service life. When a new steel strip is installed, the timer is reset to zero and starts timing, and at the same time, the expected service life of the steel strip is input and the alarm threshold is set. Since there may be a certain error accumulation in electronic devices, the timer is calibrated every once in a while to ensure its accuracy. When the timer reaches the declared service life, it is judged that the steel strip meets the scrapping technical requirements, such as Figure 4 .

[0111] 3.2 Scrapping judgment based on the number of bends of the steel strip

[0112] Install a vibration sensor 8 on the steel belt traction machine 7 to detect the start-up times of the traction machine. When the elevator traction machine starts, a specific vibration signal will be generated. This mode is different from the vibration signals when the traction machine is running or stopped normally. By monitoring the changes in these vibration signals, the start-up times of the traction machine can be indirectly judged, and then the bending times of the steel belt can be inferred.

[0113] Install a vibration sensor on the traction machine and connect it to a data acquisition system for signal acquisition. The original vibration signal may contain noise and other interference factors. Therefore, it is necessary to preprocess the signal, including steps such as filtering and noise reduction, to improve the signal quality. Extract useful features from the preprocessed signal. For judging the start of the traction machine, when the vibration signal suddenly increases from the baseline (i.e., a stable state close to zero), it can be regarded as an indication of the start of the traction machine.

[0114] First, determine a suitable threshold to distinguish the stationary state and the start state of the traction machine. The vibration signal can be recorded for a period of time when the traction machine is stationary, and its average value and standard deviation are calculated. Then, set the threshold to the average value plus 3 times the standard deviation. Real-time collect the vibration signal and continuously compare it with the set threshold. If the vibration signal exceeds the threshold and the signal has been below the threshold for a previous period of time, it is considered that the traction machine has started once.

[0115] To avoid misjudgment caused by short-term noise or interference, a "de-interference" mechanism can be introduced. Set a time window. Only when the vibration signal continuously exceeds the threshold for a time longer than this window is it considered a valid start. This can filter out short-term fluctuations and ensure the accuracy of counting. Whenever a valid start is detected, the counter is incremented by 1. When the counter data reaches the allowable simple bending times, it is judged that the steel belt meets the scrapping technical requirements, such as Figure 5 。

[0116] (4) Prediction of the remaining service life of the steel belt based on the BP neural network

[0117] By predicting the remaining service life of the elevator steel belt, it can ensure the reliable and healthy operation of the elevator and reduce the casualties caused by elevator failures. The present invention predicts the remaining service life of the steel belt through the BP neural network according to the bending times of the steel belt, the thickness of the steel belt, and the load of the elevator, such as Figure 6 。

[0118] 4.1 Input parameters

[0119] 1) Bending times of the steel belt

[0120] The number of bends of the steel belt is detected by installing a vibration sensor on the traction machine of the steel belt elevator to detect the number of starts of the traction machine. When the elevator traction machine starts, a specific vibration signal will be generated. This mode is different from the vibration signals when the traction machine is running or stopped normally. By monitoring the changes in these vibration signals, the number of starts of the traction machine can be indirectly judged, and then the number of bends of the steel belt can be inferred.

[0121] 2) Steel belt thickness

[0122] The thickness of the steel belt is obtained by two industrial cameras fixed on the left and right sides of the steel belt to obtain the side image of the steel belt. After preprocessing and edge detection, the contour of the steel belt is obtained. By calculating the vertical distance of each point on the contour, the actual thickness of the steel belt is converted according to the proportional relationship between the distance in the image and the actual thickness as the measured thickness of the steel belt.

[0123] 3) Elevator load

[0124] The load of the elevator is measured by a load sensor installed in the elevator. The sensor can sense the total weight in the elevator. When passengers enter the elevator, the sensor will detect the change in weight and send the data to the elevator control system.

[0125] Each time the elevator motor starts, the number of bends of the steel belt, the thickness, and the elevator load are recorded once. The recorded data is used as the input of the BP neural network to predict the remaining service life of the steel belt.

[0126] 4.2 Network structure

[0127] The BP neural network is a neural network that propagates errors backward. Its main feature is that the weights and biases of the network are learned through a supervised learning method, that is, the weights and biases of the network are continuously updated through the backpropagation algorithm to make the predicted output of the network as close as possible to the true value, such as Figure 7 . In this invention, the number of bends of the steel belt, the thickness of the steel belt, and the load of the elevator are selected as input features, and the remaining service life of the steel belt is used as the output target.

[0128] 1) Forward propagation

[0129] The input signal is transmitted from the input layer to the hidden layer and then to the output layer. After each layer of neurons receives the output of the previous layer, it is processed by the Sigmoid activation function to generate the output of this layer. For a neuron, its output a can be calculated by the following formula:

[0130]

[0131] In formula (10), f is the activation function, ω i is the weight, x i is the input, b is the bias, and n is the number of inputs.

[0132] 2) Calculation Error

[0133] The difference between the output of the output layer and the true value is called the error, which is used to measure the prediction performance of the network. The error is calculated using the Mean Squared Error (MSE). The formula is as follows:

[0134]

[0135] In Equation (11), y i is the true value, is the predicted value, and m is the number of samples.

[0136] 3) Backpropagation

[0137] The error propagates backward from the output layer to the input layer through the weights of the network. The chain rule is used to calculate the gradient, and the partial derivative of the error with respect to each weight is calculated. For the error gradient of the output layer:

[0138]

[0139] In Equation (12), ω oj is the weight connecting the o-th output neuron and the j-th hidden layer neuron, and z j is the weighted input of the j-th hidden layer neuron.

[0140] 4) Weight Update

[0141] According to the gradient descent algorithm, the calculated gradient is used to update the weights and biases of the network. The update rule is:

[0142]

[0143] In Equation (13), ω is the weight, η is the learning rate, is the partial derivative of the error with respect to the weight.

[0144] By continuously iterating this process until the error of the network is reduced to an acceptable threshold or a predetermined number of iterations is reached. Using the trained BP neural network, the time when the measured thickness of the steel strip is reduced to the specified value provided by the manufacturer can be predicted, and then the remaining service life of the steel strip can be predicted.

[0145] In summary, the present invention determines whether the steel strip meets the scrap technical conditions through various methods, and at the same time has a certain real-time monitoring and prediction ability. It monitors the surface defects and thickness changes of the steel strip through machine vision, and at the same time records the service time and bending times of the steel strip to determine whether the steel strip meets the scrap technical conditions; uses a BP neural network to predict the remaining service life of the steel strip. Through the present invention, it is possible to monitor the scrap technical conditions such as the surface defects and thickness of the steel strip in real time, and predict the remaining service life of the steel strip, helping maintenance personnel to make scrap judgments and life predictions for the steel strip, and ensuring the safety and reliability of the steel strip elevator operation.

[0146] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article or device including the said element.

[0147] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An elevator composite steel belt life assessment device, comprising a steel belt (5), a steel belt traction wheel (6) and a steel belt traction machine (7), characterized in that: The steel belt (5) is respectively provided with a steel belt front industrial camera (1), a steel belt right side industrial camera (2), a steel belt left side industrial camera (3) and a steel belt back side industrial camera (4) around it, and a vibration sensor (8) is provided on the top of the steel belt traction machine (7); The vibration sensor (8) determines the number of starts of the steel belt traction machine (7) by monitoring the change of the vibration signal of the motor, and further infers the number of bends of the steel belt (5).

2. The elevator composite steel belt life assessment device according to claim 1 further comprises a timer installed in the elevator control cabinet and a load sensor installed in the elevator, characterized in that: The timer records the service time of the steel belt (5) and issues an early warning when the predetermined service life of the steel belt (5) is approaching. The load sensor is used to measure the load-bearing capacity of the elevator.

3. A method for evaluating the life of an elevator composite steel belt, characterized in that: The method comprises the following steps: S1: The surface condition of the steel belt is obtained by industrial cameras installed around the steel belt, and the image is denoised, converted, and features extracted; S2: Based on the extracted geometric shape features, texture features, and edge features, a classifier is used to identify surface defects and determine whether the scrapping technical conditions have been met; S3: The extracted edge features are also used to calculate the actual thickness of the steel strip. The scrapping is determined based on whether the actual thickness is reduced to the specified value provided by the manufacturer relative to the nominal thickness. S4: By recording the service time and bending times of the steel strip, determine whether it is scrapped based on whether the allowable simple bending times and declared life are reached; S5: Based on the measured bending times of the steel belt, the actual thickness of the steel belt, and the load-bearing capacity of the elevator, the BP neural network algorithm is used to predict the time when the steel belt reaches the service data and its remaining service life.

4. The elevator composite steel belt life assessment method according to claim 3, characterized in that: In step S1, for image denoising, the mean filtering principle is adopted. By calculating the average value in the neighborhood of the pixel point, the average value of the grayscale values ​​of all pixels is used to replace the original pixel to perform noise reduction on the image. For image conversion, the weighted average method is used to convert the image to convert the color image into an image with only grayscale levels, thereby enhancing the features in the image and making the defects more obvious, which is convenient for subsequent feature extraction and analysis.

5. The elevator composite steel belt life assessment method according to claim 3, characterized in that: In step S2, the geometric shape features use two geometric parameters, area and perimeter, to describe the characteristics of the surface defects of the steel strip. The area parameter is calculated by traversing each pixel point in the defect area and adding their values ​​to obtain the total number of pixels in the defect area. The perimeter parameter is obtained by calculating the total number of pixels on the boundary of the defect area.

6. The elevator composite steel belt life assessment method according to claim 3, characterized in that: In the step S2, texture features are used to describe complex patterns on the surface of an object, and a local binary pattern is used to capture the texture.

7. The elevator composite steel belt life assessment method according to claim 3, characterized in that: In step S2, the edge feature provides information about the boundary of the object, which is convenient for identifying the shape and outline of the object. The Canny edge detection algorithm is used to select a suitable detection window size, and the gradients in the horizontal and vertical directions are calculated respectively by the Sobel operator. Then, the amplitude and direction of the gradient are calculated by the gradient, and the gradient amplitude is non-maximum suppressed to eliminate false edge information. At the same time, the edges are classified using double thresholds.

8. The elevator composite steel belt life assessment method according to claim 3, characterized in that: In step S2, the classifier uses a support vector machine (SVM). The SVM automatically detects the type and location of the surface defects of the steel strip according to the input defect feature vector through training and learning, and determines whether the defects reach the scrap technical conditions.

9. The elevator composite steel belt life assessment method according to claim 3, characterized in that: In step S3, the actual thickness of the steel strip is calculated by selecting reference points corresponding to the same height of the steel strip on both side edges of the contour as much as possible, calculating the Euclidean distance between the reference points, and converting the actual thickness of the steel strip by the proportional factor between the distance in the image and the actual thickness. Since the steel strip will deviate during the operation of the elevator, the minimum thickness is selected as the actual thickness of the steel strip, and whether the steel strip is scrapped is determined based on whether the actual thickness is reduced to the specified value provided by the manufacturer relative to the nominal thickness.

10. The elevator composite steel belt life assessment method according to claim 3, characterized in that: In step S5, each time the elevator motor is started, the bending times, thickness and elevator load of the steel belt are recorded, and the recorded data are used as input of the BP neural network to predict the remaining service life of the steel belt.

Images