An ultrasonic intelligent assessment method for bridge swivel pin damage identification

Abstract

The invention discloses an ultrasonic intelligent assessment method suitable for identifying damage to a bridge swivel pin, relates to the technical field of bridge construction, and solves the technical problem of difficulty in quickly identifying damage to a swivel pin in existing bridge swivel construction. The method comprises: an ultrasonic sensor, an ultrasonic sensor core processor, a data filter, and a pre-trained convolutional neural network deep learning operation program for analyzing and identifying the degree of damage to the pin arranged in the bridge swivel pin; ultrasonic data of the pin is obtained by an ultrasonic sensor arranged at the axis center position of the pin, the data is converted into waveform image data after wavelet denoising processing, and is imported into a pre-built and trained convolutional neural network for analysis and identification, so as to quickly and accurately obtain the damage condition of the bridge swivel pin. The invention can accurately identify the damage condition of the bridge swivel pin under the processing of the convolutional neural network, and can effectively improve the safety performance of bridge swivel construction.

Description

Technical Field

[0001] The present invention belongs to the technical field of bridge construction, and specifically relates to a swivel bridge construction, and more specifically, to an ultrasonic intelligent evaluation method suitable for identifying bridge swivel pin shaft damage. Background Art

[0002] With the rapid development of my country's economy and science and technology, transportation has made great progress as a basic condition for improving people's living standards. In recent years, a large number of bridge constructions have been put on the agenda. In bridge construction, there have been many planned bridges that need to cross existing lines and canyon projects with harsh conditions. In this case, bridge construction mostly adopts the construction form of bridge rotation. With the development of technology, the structural weight of the rotating bridge is getting heavier, and the amount of unbalanced load is getting larger. The project is also shifting to areas with relatively poor meteorological conditions such as the west. When building rotating bridges, the impact of the environment is getting greater. According to the research of existing scholars, the bridge rotation pin is often subjected to large shear force during engineering construction, which seriously endangers the normal operation of the bridge pin. It is necessary to introduce a more advanced and intelligent monitoring method for the degree of pin damage during the construction of bridge rotation.

[0003] Ultrasonic sensors have the characteristics of high frequency, short wavelength, small diffraction phenomenon, especially good directionality, and the ability to propagate in a directional manner as rays. Ultrasonic waves have great penetrating power for liquids and solids, especially in solids that are opaque to sunlight. When ultrasonic waves encounter impurities or interfaces, they will produce significant reflections to form reflected echoes. They are now widely used in non-destructive testing of concrete and various steel devices, but their application in non-destructive testing of bridge swivel pins is blank. It is necessary to introduce this technology to monitor the damage of bridge swivel pins.

[0004] The deep learning operation of convolutional neural network is a kind of artificial neural network, which has become a research hotspot in the current image recognition field. Its weight-sharing network structure makes it more similar to biological neural networks, reducing the number of weights. Convolutional neural network can directly use multi-dimensional images as the input of the neural network, avoiding the complex feature extraction and data reconstruction process in traditional recognition algorithms and having a higher recognition accuracy.

[0005] During the bridge construction process, the structural unbalanced load of the bridge structure itself and the external force on the bridge will often generate a large shear force and a large sudden bending moment on the bridge's rotating pin, causing great structural damage to the pin structure, which will seriously affect the construction accuracy and safety of the rotating bridge, and need to be quickly identified and resolved in a timely manner. Summary of the invention

[0006] The purpose of the present invention is to solve the technical problem of difficulty in quickly identifying damage to a swivel pin during existing bridge swivel construction. The present invention provides an ultrasonic intelligent assessment method suitable for identifying damage to a bridge swivel pin.

[0007] The technical solution adopted by the present invention is as follows: an ultrasonic intelligent assessment method suitable for bridge swivel pin damage identification, comprising the following steps:

[0008] S1: Arrange an ultrasonic sensor in the rotation pin of the bridge rotation construction, and connect the ultrasonic sensor to the ultrasonic core processor through a data transmission line;

[0009] S2: The ultrasonic data of the bridge swivel pin shaft obtained by the ultrasonic core processor is transmitted to the filter, and the wavelet analysis is used to perform data filtering processing on the signal denoising, and the processed ultrasonic data is converted into waveform image data;

[0010] S3: Construct a convolutional neural network for training, import the waveform image data collected in advance and processed by the ultrasonic sensor core processor into the convolutional neural network, and assign corresponding damage degree labels to the data, so that the convolutional neural network can be trained and learned, and the damage of the pin can be quickly identified according to the imported training model. By testing the accuracy of the convolutional neural network fault diagnosis, the staff will compare the convolutional neural network diagnosis results with the actual pin damage type to demonstrate the correctness of the diagnosis conclusion, meet the accuracy requirements of practical applications, and obtain a convolutional neural network model that can automatically identify pin damage and pin undamaged images;

[0011] S4: importing the waveform image data obtained in step S2 into the convolutional neural network deep learning program trained in advance in step S3 for analysis and recognition;

[0012] S5: Use the convolutional neural network model trained in step S3 to analyze and identify the waveform image data collected by the ultrasonic sensor in the bridge swivel pin and quickly obtain the damage condition of the bridge swivel pin.

[0013] In step S1, the ultrasonic sensor is arranged at the axis of the bridge swivel pin. During installation, the ultrasonic sensor data transmission line is arranged in a pre-buried corrugated pipe, and the data transmission line is wrapped with waterproof tape to prevent water from entering.

[0014] In step S2, the ultrasonic data collected by the ultrasonic sensor is converted into 224×224 waveform image data.

[0015] In step S3, the ultrasonic waveform image data of the bridge swivel pin is labeled with corresponding features, and the waveform image data is labeled as linear crack damage to the pin, network crack damage to the pin, and no damage to the pin according to the waveform.

[0016] In step S3, the ultrasonic waveform image data of the bridge swivel pin collected in advance and labeled are imported into the constructed convolutional neural network for training and learning, wherein the feature recognition of the ultrasonic waveform image data is evaluated according to the following standards: a: Linear crack damage on the bridge swivel pin has the following characteristics: the ultrasonic image defect echo amplitude is large and the peak is sharp; b: Reticular crack damage on the bridge swivel pin has the following characteristics: when the ultrasonic probe scans the defect, there are defect echoes almost all around the defect; c: The bridge swivel pin is not damaged and has the following characteristics: the ultrasonic waveform curve is flat and there is no obvious peak.

[0017] In step S3, a certain proportion of ultrasonic waveform images of linear crack damage on the bridge swivel pin, mesh crack damage on the pin and undamaged pin are used for training and learning of the convolutional neural network. The ultrasonic waveform image data set should have more than 10,000 to ensure the training accuracy of the convolutional neural network after training.

[0018] In step S3, the convolutional neural network model uses the vgg-16 network model.

[0019] In step S4, the processed waveform image data collected by the ultrasonic sensor of the bridge swivel pin is imported into the trained convolutional neural network in step S3 to perform damage identification.

[0020] In step S5, a convolutional neural network is used to quickly identify the ultrasonic waveform image data of the bridge swivel pin and obtain the damage condition of the bridge swivel pin.

[0021] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0022] The present invention can be widely used in bridge rotation construction. During the bridge rotation construction process, due to the unbalanced load of the structure and the external load, a large concentrated stress is generated at the bridge ball joint, which has a great impact on the bridge rotation construction, causing the bridge pin shaft to produce structural deformation or even brittle fracture, which will lead to serious construction safety accidents. The invention is proposed to effectively and quickly identify the above problems;

[0023] During the bridge rotation process, the pin is subjected to complex stress conditions. The ultrasonic sensor located in the middle of the pin can emit ultrasonic waves. When the ultrasonic waves hit impurities or interfaces, they will produce significant reflections to form reflected echoes. By continuously collecting ultrasonic data of the pin, the collected data can be quickly and intelligently identified for pin damage through deep learning operations performed by convolutional neural networks. Based on this function, it can be used to detect the integrity of the pin structure of the rotating bridge. If micro cracks are generated under the unbalanced stress of a bridge structure weighing tens of thousands of tons, they can also be quickly identified, which can effectively reflect the structural integrity of the pin structure in real time.

[0024] The invention has greatly improved the intelligence and safety performance of the bridge rotation technology. Based on the current method of scholars arranging sensors at the polytetrafluoroethylene slider of the bridge ball joint or at the base of the bridge, the invention can be more convenient in construction, can enhance the safety factor of the bridge rotation construction and improve the quality of the project, provide certain guidance for the rotation construction, provide a new solution for the bridge rotation construction, and effectively improve the anti-overturning ability of the bridge rotation construction structure.

[0025] In summary, the present invention can accurately identify the damage condition of the bridge rotation pin under the processing of the convolutional neural network, and has no effect on the bridge rotation construction. It has stable operation and strong anti-interference ability, and can effectively improve the safety performance of the bridge rotation construction. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The present invention will now be described by way of example with reference to the accompanying drawings, in which:

[0027] Figure 1 is an operational flow chart of the present invention;

[0028] Figure 2 It is a cross-sectional schematic diagram of the ultrasonic sensor of the bridge swivel pin of the present invention;

[0029] Figure 3 1. It is a top view schematic diagram of the ultrasonic sensor for the bridge swivel pin of the present invention;

[0030] Figure 4 It is a characteristic label classification diagram of ultrasonic waveform image data of the present invention;

[0031] Figure 5 It is a cross-sectional view of the installation of the bridge swivel pin of the present invention;

[0032] Figure 6 This is an analysis diagram of the principle of crack collection by the ultrasonic sensor of the present invention;

[0033] Figure 7 It is the operation diagram of the vgg-16 convolutional neural network used in the present invention;

[0034] Markings in the figure are: 1-bridge swivel pin, 2-ultrasonic sensor, 3-data transmission line, 4-upper ball joint, 5-lower ball joint, 6-micro crack. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0036] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. 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.

[0037] Example 1

[0038] like Figure 1-7 As shown, this embodiment provides an ultrasonic intelligent assessment method suitable for bridge swivel pin damage identification, comprising the following steps:

[0039] S1: During the bridge rotation construction, Figure 2-3 An ultrasonic sensor 2 is arranged in the bridge swivel pin 1, and the ultrasonic sensor 2 is connected to the ultrasonic core processor via a data transmission line 3;

[0040] S2: The ultrasonic data of the bridge swivel pin 1 shaft obtained by the ultrasonic core processor is transmitted to the filter, and the signal is de-noised and data filtered by wavelet analysis, and the processed ultrasonic data is converted into 224×224 waveform image data;

[0041] S3: Construct a convolutional neural network for training, import the 224×224 waveform image data collected in advance and processed by the ultrasonic sensor core processor into the convolutional neural network, and assign corresponding damage degree labels to it. Figure 4The waveform shown in the figure is assigned with waveform feature labels: a: Linear crack damage on the bridge swivel pin has the following characteristics: the ultrasonic image defect echo amplitude is large and the wave peak is sharp; b: The reticular crack damage on the bridge swivel pin has the following characteristics: when the ultrasonic probe scans the defect, there are defect echoes almost all around the defect; c: The bridge swivel pin has the following characteristics without damage: the ultrasonic waveform curve is gentle and there is no obvious wave peak; The convolutional neural network is trained and learned, and the pin damage is quickly identified based on the imported training model. By testing the accuracy of the convolutional neural network fault diagnosis, the staff compared the convolutional neural network diagnosis results with the actual pin damage type to demonstrate the correctness of the diagnosis conclusion, meet the accuracy requirements of practical applications, and obtain a convolutional neural network model that can automatically identify pin damage and pin undamaged images;

[0042] S4: importing the 224×224 waveform image data obtained in step S2 into the convolutional neural network deep learning program trained in advance in step S3 for analysis and recognition;

[0043] S5: Use the convolutional neural network model trained in step S3 to analyze and identify the waveform image data collected by the ultrasonic sensor 2 in the bridge swivel pin 1 and quickly obtain the damage condition of the bridge swivel pin 1.

[0044] The working principle of the present invention is as follows: the present invention includes an ultrasonic sensor 2, an ultrasonic sensor core processor, a data filter, and a pre-trained convolutional neural network deep learning operation program for analyzing and identifying the degree of damage to the pin shaft, which is arranged in the bridge rotation pin shaft 1; when micro cracks 6 appear on the pin shaft, data is collected by the ultrasonic sensor 2 and the damage to the pin shaft is quickly identified and judged through the operation of the data filter and the trained convolutional neural network, so that the bridge construction can be stopped immediately, and related maintenance and processing work can be carried out to prevent the risk of reduced bridge accuracy and bridge overturning caused by pin shaft breakage, thereby improving the safety factor of the rotating bridge.

[0045] Example 2

[0046] On the basis of Example 1, in step S1, the ultrasonic sensor 2 is arranged at the axial position of the bridge swivel pin 1. During installation, the data transmission line 3 of the ultrasonic sensor 2 is to be arranged in a pre-buried corrugated pipe, and the data transmission line 3 is wrapped with waterproof tape to prevent water from entering.

[0047] Example 3

[0048] On the basis of Example 1, in step S3, the ultrasonic waveform image data of the bridge swivel pin 1 is labeled with corresponding features, and the waveform image data is labeled as pin linear crack damage, pin network crack damage and pin undamaged according to the waveform.

[0049] Example 4

[0050] On the basis of Example 1, in step S3, a certain proportion of ultrasonic waveform images of linear crack damage on the bridge swivel pin 1, network crack damage on the pin and undamaged pin are used for training and learning of the convolutional neural network. The ultrasonic waveform image data set should have more than 10,000 to ensure the training accuracy of the convolutional neural network after training.

[0051] Example 5

[0052] Based on Example 1, the model of the convolutional neural network in step S3 is selected as follows Figure 7 The vgg-16 network model shown.

[0053] Example 6

[0054] On the basis of Example 1, in step S4, the waveform image data collected by the ultrasonic sensor 2 of the bridge swivel pin 1 is imported into step S3, and damage identification is performed in the trained convolutional neural network.

[0055] Example 7

[0056] On the basis of Example 1, in step S5, a convolutional neural network is used to quickly identify the ultrasonic waveform image data of the bridge swivel pin 1, and the damage condition of the bridge swivel pin 1 is obtained.

Claims

1. An ultrasonic intelligent assessment method suitable for bridge swivel pin damage identification, characterized in that: The following steps are involved: S1: Arrange an ultrasonic sensor in the rotation pin of the bridge rotation construction, and connect the ultrasonic sensor to the ultrasonic core processor through a data transmission line; S2: The ultrasonic data of the bridge swivel pin shaft obtained by the ultrasonic core processor is transmitted to the filter, and the signal is denoised by wavelet analysis to perform data filtering processing, and the processed ultrasonic data is converted into waveform image data; S3: Construct a convolutional neural network for training, import the waveform image data collected in advance and processed by the ultrasonic sensor core processor into the convolutional neural network, and assign corresponding damage degree labels to it, so that the convolutional neural network can be trained for learning, and the pin damage can be quickly identified according to the imported training model. By testing the accuracy of the convolutional neural network fault diagnosis, the staff will compare the convolutional neural network diagnosis results with the actual pin damage type to demonstrate the correctness of the diagnosis conclusion, meet the accuracy requirements of practical applications, and obtain a convolutional neural network that can automatically identify pin damage and pin undamaged images; S4: importing the waveform image data obtained in step S2 into the convolutional neural network deep learning program trained in advance in step S3 for analysis and recognition; S5: using the convolutional neural network trained in step S3, analyzing and identifying the waveform image data collected by the ultrasonic sensor in the bridge swivel pin and quickly obtaining the damage condition of the bridge swivel pin; In step S3, the ultrasonic waveform image data of the bridge swivel pin is labeled with corresponding features, and the waveform image data is labeled as linear crack damage to the pin, network crack damage to the pin, and no damage to the pin according to the waveform; In step S3, the ultrasonic waveform image data of the bridge swivel pin collected in advance and labeled are imported into the constructed convolutional neural network for training and learning, wherein the ultrasonic waveform image data feature recognition is evaluated according to the following standards: a: Linear crack damage on the bridge swivel pin has the following characteristics: The ultrasonic image defect echo amplitude is large and the wave peak is sharp; b: The mesh crack damage on the bridge swivel pin has the following characteristics: When the ultrasonic probe scans the defect, there are defect echoes almost all around the defect; c: The bridge swivel pin has the following characteristics without damage: The ultrasonic waveform curve is flat and there is no obvious wave peak; In step S3, a certain proportion of ultrasonic waveform images of linear crack damage on the bridge swivel pin, network crack damage on the pin, and undamaged pin are imported into the constructed convolutional neural network for training and learning. The ultrasonic waveform image data set should have more than 10,000 to ensure the recognition accuracy of the trained convolutional neural network.

2. The ultrasonic intelligent assessment method for bridge swivel pin damage identification according to claim 1 is characterized in that : In step S1, the ultrasonic sensor is arranged at the axis of the bridge swivel pin. During installation, the ultrasonic sensor data transmission line should be arranged in a pre-buried corrugated pipe, and the data transmission line should be wrapped with waterproof tape to prevent water from entering.

3. The ultrasonic intelligent assessment method for bridge swivel pin damage identification according to claim 1 is characterized in that : Step S2 converts the ultrasonic data collected by the ultrasonic sensor into 224×224 waveform image data.

4. The ultrasonic intelligent assessment method for bridge swivel pin damage identification according to claim 1 is characterized in that : In step S3, the convolutional neural network model uses the vgg-16 network model.

5. The ultrasonic intelligent assessment method for bridge swivel pin damage identification according to claim 1 is characterized in that In step S4, the waveform image data collected by the ultrasonic sensor of the bridge swivel pin is imported into step S3, and the pin damage is identified by the pre-trained convolutional neural network.

6. The ultrasonic intelligent assessment method for bridge swivel pin damage identification according to claim 1 is characterized in that : In step S5, a convolutional neural network is used to quickly identify the ultrasonic waveform image data of the bridge swivel pin and obtain the damage condition of the bridge swivel pin.

Images