Elevator steel belt surface damage detection method and system based on deep learning

By developing a deep learning-based method for detecting surface damage to elevator steel strips, this method solves the problems of time-consuming and labor-intensive manual inspection and complex non-destructive testing in existing technologies. It achieves efficient and accurate detection of elevator steel strip damage, thus ensuring elevator safety.

CN119027373APending Publication Date: 2024-11-26CHONGQING SPECIAL EQUIP INSPECTION & RES INST
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411002187.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-25
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

Current methods for detecting surface damage to elevator steel belts mainly rely on manual visual inspection, which is labor-intensive, prone to errors, cannot be performed in real time, and lacks unified standards. Existing non-destructive testing technologies are complex and costly, and cannot accurately analyze the type and extent of damage.

Method used

A deep learning-based approach is adopted to automatically detect surface damage on elevator steel strips by acquiring, preprocessing, and dividing the dataset and training a damage detection model, and using the YOLOv3 model and its improved structure.

Benefits of technology

It has enabled automated detection of surface damage to elevator steel belts, improving detection efficiency and accuracy, reducing labor costs, and ensuring the safe and reliable operation of elevators.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119027373A_ABST
    Figure CN119027373A_ABST
Patent Text Reader

Abstract

The invention discloses an elevator steel belt surface damage detection method and system based on deep learning, and relates to the technical field of elevator steel belt surface damage detection. The method includes: acquiring elevator steel belt data; preprocessing the acquired elevator steel belt data; dividing the preprocessed elevator steel belt data into a training set and a test set; inputting the training set into a damage detection model, updating network weight parameters through a loss function, and obtaining a trained damage detection model after a plurality of times of training; and inputting the test set into the trained damage detection model, and evaluating an elevator steel belt damage detection identification result. According to the invention, automatic detection can be realized, the detection efficiency is greatly improved, and the labor cost is reduced; the detection efficiency, the detection speed and the detection accuracy of the elevator steel belt surface damage are improved, and safe and reliable operation of an elevator is guaranteed.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of elevator steel belt surface damage detection, and in particular to a method and system for detecting elevator steel belt surface damage based on deep learning. Background Art

[0002] The elevator steel belt is a new type of traction suspension part, mainly composed of an internal metal steel wire rope and an externally coated polyurethane material. The polyurethane material coated on the outside of the elevator steel belt can protect the steel wire rope core from rust and wear damage. During the operation of the elevator, the polyurethane material coated on the surface of the elevator steel belt is prone to wear damage when it comes into contact with the traction wheel and guide wheel, and surface cracks and exposed steel wire damage occur under external impact. If the surface damage of the elevator steel belt is not discovered in time, it will further affect the safety and reliability of the elevator.

[0003] The safety inspection of elevator steel belts is a crucial part of elevator safety inspection. At present, the surface damage detection of elevator steel belts is mainly based on regular inspection by human eyes, but the use of this detection method will seriously consume manpower. Human eye inspection is generally performed by staff to conduct visual inspections at relevant places from time to time. This method cannot detect the surface damage of steel belts in real time and has a relatively high detection error. The current non-destructive testing technologies for elevator steel belts in the industrial field, such as electromagnetic detection technology based on metal materials, life prediction value method, resistance detection method, etc., generally have problems such as high price, complex operation, and high professional knowledge requirements. In addition, there is no unified detection standard for residual strength of domestic elevator steel belts. The surface damage of elevator steel belts is complex and diverse. The existing non-destructive testing devices for elevator steel belts cannot conduct qualitative and quantitative analysis of the damage type and degree of damage to the steel belts.

[0004] Therefore, it is an urgent problem for technical personnel in this field to propose a method and system for detecting surface damage of elevator steel belts based on deep learning to solve the difficulties existing in the prior art. Summary of the invention

[0005] In view of this, the present invention provides a method and system for detecting surface damage of elevator steel belts based on deep learning, which can realize automated detection, greatly improve detection efficiency, and reduce labor costs; it improves the detection efficiency, detection speed, and detection accuracy of surface damage of elevator steel belts, which is conducive to ensuring safe and reliable operation of elevators.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] A method for detecting surface damage of an elevator steel belt based on deep learning comprises the following steps:

[0008] S1. Get data: get elevator steel belt data;

[0009] S2. Data preprocessing: preprocessing the acquired elevator steel belt data;

[0010] S3. Data division: divide the preprocessed elevator steel belt data into a training set and a test set;

[0011] S4. Model training: input the training set into the damage detection model, update the network weight parameters through the loss function, and obtain the trained damage detection model after several trainings;

[0012] S5. Detection and identification: Input the test set into the trained damage detection model to evaluate the elevator steel belt damage detection and identification results.

[0013] Optionally, in S1, an industrial high-speed camera is used to obtain image data of the steel strip surface coating layer.

[0014] Optionally, in S2, contrast enhancement, image smoothing, image segmentation and extraction, and image defogging are performed on the acquired elevator steel belt data.

[0015] Optionally, the damage detection model in S4 adopts a YOLOv3 model, including a backbone network, a bottleneck layer, and a feature fusion structure layer;

[0016] The backbone network is the YOLOv3 backbone network Darknet53 with the SPP structure added;

[0017] The bottleneck layer is the FPN-light structure obtained by simplifying the original FPN layer of YOLOv3;

[0018] The feature fusion structure layer selects the N1 module and N2 module of Resnet50 as parallel auxiliary structures, and fuses them with the backbone network Darknet53 to obtain the feature fusion structure of the deep network.

[0019] Optionally, in S4, the training set is input into the damage detection model, and the network weight parameters are updated through the loss function. After several trainings, the specific content of the trained damage detection model is:

[0020] The network randomly initializes weights so that the initialized values ​​follow a Gaussian normal distribution;

[0021] The damage detection model inputs the training set data and outputs three feature maps of different scales, and uses these feature maps to obtain the prediction box information;

[0022] Compare the predicted box with the real box and calculate the loss error;

[0023] When the number of iterations is less than epoch = 120, the Adam optimization algorithm is used to update the weights until the loss value is lower than the set threshold or the number of iterations is greater than epoch, the training is terminated, and a trained damage detection model is obtained.

[0024] A detection system for surface damage of elevator steel strips based on deep learning, applying any of the above-mentioned detection methods for surface damage of elevator steel strips based on deep learning, comprising: a data acquisition module, a data preprocessing module, a data division module, a model training module and a detection and recognition module;

[0025] A data acquisition module is connected to the input end of the data preprocessing module and is used to acquire elevator steel belt data;

[0026] A data preprocessing module, connected to the input end of the data partitioning module, for preprocessing the acquired elevator steel belt data;

[0027] A data partitioning module, connected to the input end of the model training module, for partitioning the preprocessed elevator steel belt data into a training set and a test set;

[0028] The model training module is connected to the input end of the detection and recognition module, and is used to input the training set into the damage detection model, update the network weight parameters through the loss function, and obtain the trained damage detection model after several trainings;

[0029] The detection and recognition module is connected to the output end of the model training module and is used to input the test set into the trained damage detection model to evaluate the elevator steel belt damage detection and recognition results.

[0030] It can be seen from the above technical solution that compared with the prior art, the present invention provides a method and system for detecting surface damage of elevator steel belts based on deep learning, which has the following beneficial effects: the present invention can realize automated detection, greatly improves detection efficiency, and reduces labor costs; improves the detection efficiency, detection speed, and detection accuracy of surface damage of elevator steel belts, which is conducive to ensuring safe and reliable operation of elevators. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0032] Figure 1 A flow chart of a method for detecting surface damage of elevator steel strips based on deep learning provided by the present invention;

[0033] Figure 2 A structural block diagram of a deep learning-based elevator steel belt surface damage detection system provided by the present invention. DETAILED DESCRIPTION

[0034] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0035] Reference Figure 1 As shown, the present invention discloses a method for detecting surface damage of an elevator steel belt based on deep learning, comprising the following steps:

[0036] S1. Get data: Get elevator steel belt data;

[0037] S2. Data preprocessing: preprocessing the acquired elevator steel belt data;

[0038] S3. Data division: divide the preprocessed elevator steel belt data into a training set and a test set;

[0039] S4. Model training: input the training set into the damage detection model, update the network weight parameters through the loss function, and obtain the trained damage detection model after several trainings;

[0040] S5. Detection and identification: Input the test set into the trained damage detection model to evaluate the elevator steel belt damage detection and identification results.

[0041] Furthermore, in S1, an industrial high-speed camera is used to obtain image data of the surface coating layer of the steel strip.

[0042] Furthermore, in S2, the acquired elevator steel belt data is subjected to contrast enhancement, image smoothing, image segmentation and extraction, and image defogging.

[0043] Specifically, the images of the steel strip surface coating acquired by the industrial high-speed camera are subjected to image preprocessing such as image contrast enhancement, image smoothing, and image segmentation and extraction. The surface damage images of the elevator steel strip collected in a humid and foggy environment are processed by Retinex defogging and dark channel prior defogging. The surface damage images collected in a low-light environment are processed by a method based on image γ correction.

[0044] Furthermore, the damage detection model in S4 adopts the YOLOv3 model, which includes a backbone network, a bottleneck layer, and a feature fusion structure layer;

[0045] The backbone network is the YOLOv3 backbone network Darknet53 with the SPP structure added;

[0046] The bottleneck layer is the FPN-light structure obtained by simplifying the original FPN layer of YOLOv3;

[0047] The feature fusion structure layer selects the N1 module and N2 module of Resnet50 as parallel auxiliary structures, and fuses them with the backbone network Darknet53 to obtain the feature fusion structure of the deep network.

[0048] Specifically, the specific content of the feature fusion structure layer is: the output features of the residual module C2 of the backbone network Darknet53 are input into the auxiliary network N1 and the residual module C3 of the backbone network Darknet53; the output features of the auxiliary network N1 are output to the auxiliary network N2, and the output features of the auxiliary network N1 are upsampled at the same time; the upsampled features of the auxiliary network N1 are fused with the output features of the residual module C3 of the backbone network Darknet53 in an accumulated form and input into the residual module C4 of the backbone network Darknet53; the output features of the auxiliary network N2 are upsampled, and then fused with the output features of the residual module C4 in the backbone network Darknet53 in an accumulated form and input into the residual module C5 of the backbone network Darknet53; the output features of the residual module C5 of the backbone network Darknet53 are sequentially sent to the SPP layer and the FPN-light bottleneck layer.

[0049] Furthermore, in S4, the training set is input into the damage detection model, and the network weight parameters are updated through the loss function. After several trainings, the specific content of the trained damage detection model is obtained as follows:

[0050] The network randomly initializes weights so that the initialized values ​​follow a Gaussian normal distribution;

[0051] The damage detection model inputs the training set data and outputs three feature maps of different scales, and uses these feature maps to obtain the prediction box information;

[0052] Compare the predicted box with the real box and calculate the loss error;

[0053] When the number of iterations is less than epoch = 120, the Adam optimization algorithm is used to update the weights until the loss value is lower than the set threshold or the number of iterations is greater than epoch, the training is terminated, and a trained damage detection model is obtained.

[0054] In a specific embodiment, 800 examples of elevator steel belt surface image data are used in advance, and the 800 images are pre-processed by contrast enhancement, image smoothing, image segmentation extraction, image defogging, etc. 60% of the images are randomly selected as training sets, and 40% of the images are used as test sets; the damage detection model is trained to reduce the loss function value and update the network weights. After continuous iterative learning, a trained damage detection model is obtained. The surface of the elevator steel belt to be detected is detected and identified through rapid modeling and testing.

[0055] and Figure 1 Corresponding to the method described above, the embodiment of the present invention also provides a detection system for elevator steel belt surface damage based on deep learning, which is used to Figure 1 The specific implementation of the method is shown in the following diagram: Figure 2 As shown, it specifically includes: data acquisition module, data preprocessing module, data partitioning module, model training module and detection and recognition module;

[0056] A data acquisition module is connected to the input end of the data preprocessing module and is used to acquire elevator steel belt data;

[0057] A data preprocessing module, connected to the input end of the data partitioning module, for preprocessing the acquired elevator steel belt data;

[0058] A data partitioning module, connected to the input end of the model training module, for partitioning the preprocessed elevator steel belt data into a training set and a test set;

[0059] The model training module is connected to the input end of the detection and recognition module, and is used to input the training set into the damage detection model, update the network weight parameters through the loss function, and obtain the trained damage detection model after several trainings;

[0060] The detection and recognition module is connected to the output end of the model training module and is used to input the test set into the trained damage detection model to evaluate the elevator steel belt damage detection and recognition results.

[0061] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0062] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for detecting surface damage of elevator steel strip based on deep learning, characterized in that: The following steps are involved: S1. Get data: get elevator steel belt data; S2. Data preprocessing: preprocessing the acquired elevator steel belt data; S3. Data division: divide the preprocessed elevator steel belt data into a training set and a test set; S4. Model training: input the training set into the damage detection model, update the network weight parameters through the loss function, and obtain the trained damage detection model after several trainings; S5. Detection and identification: Input the test set into the trained damage detection model to evaluate the elevator steel belt damage detection and identification results.

2. The method for detecting surface damage of elevator steel strip based on deep learning according to claim 1 is characterized in that: In S1, an industrial high-speed camera is used to obtain image data of the steel strip surface coating layer.

3. The method for detecting surface damage of elevator steel strip based on deep learning according to claim 1 is characterized in that: In S2, the acquired elevator steel belt data is subjected to contrast enhancement, image smoothing, image segmentation and extraction, and image defogging.

4. The method for detecting surface damage of elevator steel strip based on deep learning according to claim 1, characterized in that: The damage detection model in S4 adopts the YOLOv3 model, which includes the backbone network, bottleneck layer and feature fusion structure layer; The backbone network is the YOLOv3 backbone network Darknet53 with the SPP structure added; The bottleneck layer is the FPN-light structure obtained by simplifying the original FPN layer of YOLOv3; The feature fusion structure layer selects the N1 module and N2 module of Resnet50 as parallel auxiliary structures, and fuses them with the backbone network Darknet53 to obtain the feature fusion structure of the deep network.

5. The method for detecting surface damage of elevator steel strip based on deep learning according to claim 1, characterized in that: In S4, the training set is input into the damage detection model, and the network weight parameters are updated through the loss function. After several trainings, the specific content of the trained damage detection model is: The network randomly initializes weights so that the initialized values ​​follow a Gaussian normal distribution; The damage detection model inputs the training set data and outputs three feature maps of different scales, and uses these feature maps to obtain the prediction box information; Compare the predicted box with the real box and calculate the loss error; When the number of iterations is less than epoch = 120, the Adam optimization algorithm is used to update the weights until the loss value is lower than the set threshold or the number of iterations is greater than epoch, the training is terminated, and a trained damage detection model is obtained.

6. A deep learning-based elevator steel belt surface damage detection system, characterized in that: A method for detecting surface damage of an elevator steel belt based on deep learning according to any one of claims 1 to 5, comprising: a data acquisition module, a data preprocessing module, a data division module, a model training module and a detection and recognition module; A data acquisition module is connected to the input end of the data preprocessing module and is used to acquire elevator steel belt data; A data preprocessing module, connected to the input end of the data partitioning module, for preprocessing the acquired elevator steel belt data; A data partitioning module, connected to the input end of the model training module, for partitioning the preprocessed elevator steel belt data into a training set and a test set; The model training module is connected to the input end of the detection and recognition module, and is used to input the training set into the damage detection model, update the network weight parameters through the loss function, and obtain the trained damage detection model after several trainings; The detection and recognition module is connected to the output end of the model training module and is used to input the test set into the trained damage detection model to evaluate the elevator steel belt damage detection and recognition results.