Concrete bridge prestressed tendon damage detection system

Abstract

The invention discloses a concrete bridge prestressed tendon damage detection system which comprises a magnetic detection unit, a signal processing unit and a damage detection unit. The magnetic detection unit is used for collecting a leakage magnetic field signal of an area to be detected; the signal processing unit is used for performing data processing on the acquired leakage magnetic field signal to obtain processed data; and the damage detection unit is used for carrying out feature extraction on the processed data to obtain feature parameters, inputting the feature parameters into a neural network model to carry out network model training, and outputting a damage detection result by utilizing the trained network model. According to the method, the damage degree of the prestressed tendon can be accurately and quantitatively evaluated, and technical support is provided for evaluation of corrosion and fatigue damage of the bridge prestressed tendon.

Description

Technical Field

[0001] The present invention relates to the field of concrete bridges, and particularly to a detection system for prestressed tendon damage of concrete bridges. Background Art

[0002] The corrosion and fatigue problems of prestressed tendons in concrete bridges have always existed. The cross-sectional loss of prestressed tendons caused by corrosion and the cracks or broken wires of prestressed tendons caused by fatigue may lead to brittle failure of the tendon bundle under the stress state and significantly reduce the service life of the structure. Therefore, effectively detecting the damage condition of prestressed tendons is crucial for the safe operation of bridges.

[0003] The magnetic flux leakage non-destructive testing technology has the advantages of high detection sensitivity, fast detection speed, low cost, simple operation, etc. This technology has been widely used in the detection of damage such as holes in the pipeline wall, corrosion pits, cracks, and welding. However, due to limited magnetization ability, lack of interpretation of complex signals, and testing means to minimize interference signals, its application in detecting the corrosion damage of internal prestressed tendons in concrete bridge structures is restricted, and there is a problem that the corrosion damage of bridge prestressed tendons cannot be accurately and comprehensively evaluated.

[0004] Therefore, to solve the above problems, a detection system for prestressed tendon damage of concrete bridges is needed, which can accurately quantify and evaluate the damage degree of prestressed tendons and provide technical support for the evaluation of corrosion damage of bridge prestressed tendons. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to overcome the defects in the prior art and provide a detection system for prestressed tendon damage of concrete bridges, which can accurately quantify and evaluate the damage degree of prestressed tendons and provide technical support for the evaluation of corrosion and other damages of bridge prestressed tendons.

[0006] The detection system for prestressed tendon damage of the present invention's concrete bridge includes a magnetic detection unit, a signal processing unit, and a damage detection unit;

[0007] The magnetic detection unit is used to collect the magnetic flux leakage signals in the area to be detected;

[0008] The signal processing unit is used to process the collected magnetic flux leakage signals to obtain the processed data;

[0009] The damage detection unit is used to extract features from the processed data to obtain feature parameters, input the feature parameters into a neural network model for training of the network model, and use the trained network model to output the damage detection result.

[0010] Furthermore, the magnetic detection unit includes an excitation device, a magnetic sensor, and an auxiliary movement device;

[0011] The excitation device uses a permanent magnet as the magnetic source, and forms a magnetic circuit through the permanent magnet, the armature and the prestressing tendon;

[0012] The magnetic sensor is used to collect the leakage magnetic field generated at the damaged part of the prestressing tendon;

[0013] The auxiliary motion device is used to achieve precise positioning and speed control in any three-dimensional space position, and drive the excitation device and the magnetic sensor to perform detection on the specified area to be detected according to the planned path.

[0014] Furthermore, the magnetic sensor adopts an array type; the array design arranges the magnetic sensors in two rows in the X and Y directions, with n magnetic sensors in each row, responsible for detecting the X and Z magnetic field components; among them, the moving direction along the beam body is the X direction, the direction perpendicular to the X direction in the horizontal plane where the beam body is located is the Y direction, and the direction perpendicular to the plane where the X direction and the Y direction are located is the Z direction.

[0015] Furthermore, 2 neodymium iron boron permanent magnets of N35 and above grades are selected, and the magnet size is determined according to the excitation requirements; a rectangular armature made of Q235 material is processed into corresponding dimensions, and a magnet is arranged at each end of the armature, and together with the two magnets, a "U" - shaped structure is formed to guide the magnetic lines of force and make the magnetic lines of force form a closed magnetic circuit.

[0016] Furthermore, data processing is performed on the collected leakage magnetic field signals, specifically including:

[0017] Removing the DC component and part of the noise of the leakage magnetic field signal, and performing operational amplification processing on the weak voltage signal.

[0018] Furthermore, removing the DC component of the leakage magnetic field signal specifically includes:

[0019] Eliminating the baseline in the leakage magnetic signal according to the following formula:

[0020]

[0021] Among them, S(i) is the leakage magnetic signal after baseline elimination; x i is the current data; n is the total number of collected data;

[0022] Let the average value of the peak - valley values be Taking the peak - valley value V a obtained from the a - th test as the reference, the corrected data S′(i) for S(i) is:

[0023]

[0024] Among them, m is the total number of tests.

[0025] Furthermore, the peak-valley values of the magnetic flux leakage signal are extracted according to the following steps:

[0026] S1. Set the peak search flag to flag = 1. When it is 1, search for the peak in the detection data; when it is 0, search for the valley in the detection data.

[0027] S2. Set the initial value of the peak Mx to -inf, the initial value of the valley Mn to inf, and the threshold of the peak-valley value to delta.

[0028] S3. The current value of the detection data is Di. Determine whether the current value Di is greater than the peak Mx. If true, set Mx = Di; otherwise, Mx remains unchanged.

[0029] S4. Determine whether the current value Di is less than the valley Mn. If true, set Mn = Di; otherwise, Mn remains unchanged.

[0030] S5. If flag is 1, determine whether Di is less than Mx - delta. If true, set Mn = Di; otherwise, Mn remains unchanged, and modify flag to 0.

[0031] S6. If flag is 0, determine whether Di is greater than Mn + delta. If true, set Mx = Di; otherwise, Mx remains unchanged, and modify flag to 1.

[0032] S7. Determine whether the current data is the last data. If not, obtain the next data and repeat S3 - S6 until all data detection and judgment are completed.

[0033] Furthermore, the characteristic parameters include the peak-valley difference of the axial signal, the peak span of the axial signal, the peak-valley difference of the normal signal, the peak-valley spacing of the normal signal, and the lift-off value of the magnetic sensor.

[0034] Furthermore, input the characteristic parameters into the neural network model for training of the network model, specifically including:

[0035] Select a suitable neural network model, set the activation function of the neural network to Sigmoid, the training error to 0.01, the learning rate to 0.01, the number of nodes in the hidden layer to 10, the number of nodes in the input layer to 5, and the number of nodes in the output layer to 2.

[0036] Obtain the characteristic parameters under several working conditions to form sample data; use 90% of the sample data as training samples, and the remaining 10% of the sample data as test samples.

[0037] The beneficial effects of the present invention are as follows: A prestressed tendon damage detection system for concrete bridges disclosed by the present invention constructs a magnetic detection unit and applies the magnetic detection unit to damage detection. By analyzing the characteristics of magnetic flux leakage signals under different corrosion conditions and the influence factors of stirrups, magnetic signal characteristic values are extracted to form damage characteristic samples such as corrosion and fatigue, and a prestressed tendon damage quantification neural network model is established, thereby performing quantitative analysis of magnetic flux leakage signals. It can accurately and comprehensively evaluate the damage degree of bridge prestressed tendons, providing technical support for the evaluation of damage such as corrosion and fatigue of bridge prestressed tendons. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The present invention will be further described below in conjunction with the drawings and embodiments:

[0039] Figure 1 It is a schematic diagram of the principle of the corrosion damage detection system of the present invention;

[0040] Figure 2 It is a schematic diagram of the structural design of the magnetic detection unit of the present invention;

[0041] Figure 3 It is a schematic diagram of the baseline estimation of the magnetic flux leakage signal of the present invention;

[0042] Figure 4 It is a schematic diagram of the peak-valley value detection of the magnetic flux leakage signal of the present invention;

[0043] Figure 5 It is a schematic diagram of the neural network structure of the present invention;

[0044] Figure 6 It is a schematic diagram of the axial waveform signal in the magnetic flux leakage signal characteristic parameters of the present invention;

[0045] Figure 7 It is a schematic diagram of the normal waveform signal in the magnetic flux leakage signal characteristic parameters of the present invention;

[0046] Figure 8 It is a schematic diagram of the comparison of corrosion rate prediction results of the present invention;

[0047] Figure 9 It is a schematic diagram of the comparison of corrosion width prediction results of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] The following further describes the present invention in conjunction with the accompanying drawings of the specification, as shown in the figure:

[0049] This embodiment discloses a prestressed tendon damage detection system for concrete bridges, including a magnetic detection unit, a signal processing unit, and a damage detection unit;

[0050] The magnetic detection unit is used to collect the magnetic flux leakage field signal of the area to be detected;

[0051] The signal processing unit is used to process the collected magnetic leakage field signals to obtain processed data;

[0052] The damage detection unit is used to extract features from the processed data to obtain feature parameters, input the feature parameters into a neural network model for training of the network model, and use the trained network model to output a damage detection result.

[0053] In this embodiment, the prestressed tendon belongs to a ferromagnetic material with a relatively high magnetic permeability, and the concrete cover is a non-ferromagnetic material. When the prestressed tendon is locally magnetized to the saturation state, the magnetic induction lines in its damaged area will be distorted, resulting in partial magnetic induction lines leaking to the metal surface, thus forming a magnetic leakage field. The magnetic sensor collects the magnetic leakage field caused by the damage. The magnitude of this magnetic leakage field is related to the magnetic induction intensity and the degree of damage after the prestressed tendon is magnetized. By processing and analyzing the collected signals, the degree of damage of the prestressed tendon can be judged.

[0054] As Figure 2 shown, the magnetic detection unit includes an excitation device, a magnetic sensor, and an auxiliary motion device; the excitation device uses a permanent magnet as a magnetic source, and forms a magnetic circuit through the permanent magnet, the armature, and the prestressed tendon; the magnetic sensor is used to collect the magnetic leakage field generated at the damaged part of the prestressed tendon; the auxiliary motion device is used to achieve precise positioning and speed control in any three-dimensional space position, and drive the excitation device and the magnetic sensor to the designated area to be inspected along the planned path for detection.

[0055] Among them, in order to achieve precise positioning and speed control in any three-dimensional space position, an existing three-axis motion control device composed of a controller, a motor driver, sensors, mechanical components, and a host computer can be used to achieve auxiliary motion; control the start and stop in the X, Y, and Z directions, as well as control the speed and displacement amount, and set the positive and negative limit values in the moving directions of the X and Z axes to achieve automatic recognition control of the moving position and automatic calibration of the initial position, so as to achieve the detection and identification of the damage of the prestressed tendon.

[0056] By using a permanent magnet as the magnetic source of the excitation device to form a stable magnetic circuit, the detection sensitivity of the magnetic leakage field signal is improved; the magnetic sensor can accurately collect the magnetic leakage field signal at the damaged part of the prestressed tendon to achieve precise detection of the damage; the auxiliary motion device can accurately control the positioning and movement of the magnetic detection unit in three-dimensional space, improve the comprehensiveness and flexibility of the detection, and ensure efficient and reliable detection of the damage of the prestressed tendon in complex concrete bridge structures.

[0057] In this embodiment, the magnetic sensor adopts an array type; the sensor probes adopt an array type, which can increase the detection coverage range and obtain richer prestressed tendon damage information. Among them, in the array design, the magnetic sensors are arranged in two rows in the X and Y directions, with n magnetic sensors in each row, responsible for detecting the X and Z (Bx, Bz) magnetic field components; among them, the moving direction along the beam body is the X direction, the direction perpendicular to the X direction in the horizontal plane where the beam body is located is the Y direction, and the direction perpendicular to the plane where the X direction and the Y direction are located is the Z direction.

[0058] Through multiple Hall sensors arranged in the X and Y directions, the leakage magnetic field signals at multiple spatial points can be collected simultaneously, improving the comprehensiveness of data acquisition and the detection efficiency; detecting the Bx and Bz magnetic field components separately in rows can more accurately reflect the damage characteristics of the prestressed tendons and enhance the damage recognition ability; at the same time, the array design can reduce the influence of the error of a single sensor and improve the detection accuracy.

[0059] In this embodiment, 2 NdFeB permanent magnets with a magnetic property of N35 or above are selected, and steel plate materials with a relatively high magnetic permeability such as Q235 are selected to process and manufacture a rectangular armature with corresponding dimensions. The sizes of the magnet and the armature are determined according to the device design and excitation requirements. For example, the size of the magnet can be 200mm×100mm×50mm; the size of the corresponding armature can be 350mm×200mm×10mm; a magnet is arranged at each end of the armature to form a "U" - shaped structure with the two magnets, guiding the magnetic force lines, reducing the magnetic resistance, making the magnetic force lines form a closed magnetic circuit, and reducing the influence of the stray magnetic field on the sensor.

[0060] The NdFeB permanent magnet with high magnetic performance provides a stable and strong magnetic field, ensuring sufficient magnetization of the prestressed tendons and improving the sensitivity of magnetic flux leakage detection; the reasonably designed rectangular armature is made of steel plate materials with high magnetic permeability such as Q235, effectively guiding the magnetic force lines to form a stable closed magnetic circuit, improving the magnetic field utilization rate; through the above settings, the interference of the stray magnetic field on the magnetic sensor can also be reduced, improving the quality of the magnetic flux leakage signal, thereby enhancing the detection accuracy and reliability of the prestressed tendon damage.

[0061] In this embodiment, data processing is performed on the collected leakage magnetic field signals, which specifically includes:

[0062] Removing the DC component and part of the noise of the leakage magnetic field signal, and performing operational amplification processing on the weak voltage signal.

[0063] By removing the DC component and part of the noise, the environmental interference is effectively reduced, the purity of the signal is improved, and thus the identifiability of the damage characteristics of the prestressed tendon is enhanced. By performing operational amplification on the weak voltage signal, the signal strength is increased, enabling accurate capture of even subtle magnetic flux leakage changes and improving the detection sensitivity. Moreover, it can optimize the quality of the input signal, provide more accurate basic data for subsequent feature extraction and neural network model training, and thus improve the accuracy and reliability of the overall corrosion detection.

[0064] In this embodiment, during the magnetic flux leakage detection process, due to different liftoff values and inconsistent excitation intensities of the array magnetic sensors, different DC biases and different AC amplitudes are output from different channels of the sensors. The output DC low-frequency fluctuation caused by factors such as liftoff value and uneven excitation is the baseline of the magnetic flux leakage signal. The segmented averaging method is used to eliminate the baseline in the magnetic flux leakage signal, and the number of segmented data is the number of sampled data within one cycle of the strand wave; as Figure 3 shown, the fluctuation of the liftoff value will cause the fluctuation of the excitation magnetic field strength, and the baseline of the magnetic flux leakage signal is the result of this fluctuation. Removing the DC component of the magnetic flux leakage field signal specifically includes:

[0065] Eliminating the baseline in the magnetic flux leakage signal according to the following formula:

[0066]

[0067] where S(i) is the magnetic flux leakage signal after baseline elimination; x i is the current data; n is the total number of collected data;

[0068] Figure 4 is the peak-valley value (the peak-valley value is the difference between the peak value and the valley value, also known as the extreme difference) detection result. Let the average value of the peak-valley value be Taking the peak-valley value V a obtained from the a-th test as the reference, the data S(i) is corrected, and the corrected data S′(i) is:

[0069]

[0070] where m is the total number of tests.

[0071] Among them, the peak-valley value of the magnetic flux leakage signal is extracted according to the following steps:

[0072] S1. Set the peak search flag to flag = 1. When it is 1, search for the peak value in the detection data; when it is 0, search for the valley value in the detection data.

[0073] S2. Set the initial value of the peak value Mx to -inf, the initial value of the valley value Mn to inf, and the threshold of the peak-valley value to delta.

[0074] S3. The current value of the detected data is Di. Determine whether the current value Di is greater than the peak value Mx. If true, let Mx = Di; otherwise, Mx remains unchanged.

[0075] S4. Determine whether the current value Di is less than the valley value Mn. If true, let Mn = Di; otherwise, Mn remains unchanged.

[0076] S5. If flag is 1, determine whether Di is less than Mx – delta. If true, let Mn = Di; otherwise, Mn remains unchanged, and modify flag to 0.

[0077] S6. If flag is 0, determine whether Di is greater than Mn + delta. If true, let Mx = Di; otherwise, Mx remains unchanged, and modify flag to 1.

[0078] S7. Determine whether the current data is the last data. If not, obtain the next data and repeat S3 - S6 until all data detection and judgment are completed.

[0079] Further, taking the peak - valley value of a certain test as a reference, use the ratio of the average peak - valley value obtained from other tests to the reference peak - valley value to correct the fluctuations of the measured peak - valley values in each test, and normalize the fluctuation range of the peak - valley values.

[0080] In this embodiment, (1×7) 1860s - grade prestressed steel strands are used as the test objects, and 17 steel strand specimens with a length of 1.5 m and a diameter of 15.2 mm are prepared. The corrosion area of the specimens uses a 5% NaCl solution and a constant current of 0.6 A. The corrosion time is determined according to calculations for fixed - point accelerated electrochemical corrosion. Considering that concrete is a non - magnetic material, for convenient test comparison, a wooden model beam with dimensions of 1.2 m × 0.4 m × 0.18 m is used for testing. The prestressed tendons are placed in the hole positions of the model beam, and a uniform scanning detection is carried out at a speed of 20 mm / s along the length direction of the prestressed tendons. By analyzing the influence laws of the corrosion degree, corrosion width, and stirrups on the characteristics of the magnetic leakage signals, magnetic signal characteristic values are extracted; as Figure 6 、 7 shown, the magnetic signal characteristic values include the peak - valley difference V x of the axial signal, the peak span L x of the axial signal, and the peak - valley difference V z of the normal signal, the peak - valley spacing L z of the normal signal; the characteristic parameters include the magnetic signal characteristic values and the lift - off value of the magnetic sensor.

[0081] By setting characteristic parameters, the variation characteristics of the magnetic leakage signal can be comprehensively reflected, and the sensitivity to the damage of prestressed tendons can be improved; the peak-valley difference and peak span can characterize the corrosion degree, the peak-valley distance can be used to analyze the spatial distribution of damage, and the lift-off value can characterize the influence of the change in the sensor position during the detection process; these characteristic parameters can enhance the input quality of the neural network model and improve the accuracy and stability of corrosion detection.

[0082] In this embodiment, since there is a close relationship between the corrosion damage of prestressed tendons and the magnetic leakage characteristic quantities, and various factors affect each other, any change in a factor will cause a change in the characteristic quantity. Therefore, as Figure 5 shown, the data is input into the neural network to establish a non-linear mapping model to learn the relationship between them.

[0083] Among them, inputting the characteristic parameters into the neural network model for training of the network model specifically includes:

[0084] Select a suitable neural network model. For example, use the BP neural network as the neural network model, set the activation function of the neural network to Sigmoid, the training error to 0.01, and the learning rate to 0.01; considering the stability of the network, the number of nodes between networks should not differ too much. The number of nodes in the hidden layer is 10, the number of nodes in the input layer is 5, and the number of nodes in the output layer is 2;

[0085] Through the design of various bridge test models, such as different corrosion rates, corrosion widths, stirrup designs, and prestress values of prestressed tendons, the characteristic parameters of the damage of prestressed tendons under several working conditions are obtained to form sample data; to ensure the accuracy of network training and avoid the overlap of training samples and test samples, 90% of the sample data is used as training samples, and the remaining 10% of the sample data is used as test samples.

[0086] After about 20 iterations of network training, the set error target requirements are met, and good output results are obtained, as Figure 8 、 9 shown.

[0087] The test output results of the network model are shown in Table 1:

[0088] Table 1

[0089]

[0090] As can be seen from Table 1, when the corrosion rate is relatively large, the analysis results are closer to the true values. When the corrosion rate is small, due to the many influencing factors in magnetic flux leakage detection, the error between the analysis results and the true values is relatively large. When the corrosion rate is 10%, the prediction accuracy of the corrosion rate is 78.4%, and the maximum error of the corrosion width is 77.77%. Generally speaking, the average prediction accuracy of the corrosion rate reaches 93.58%, and the average prediction accuracy of the corrosion width reaches 90.3%.

[0091] Using the network model trained by the present invention, the accuracy of predicting the corrosion rate and the corrosion width is relatively high; the present invention can effectively reduce the influence of external factors on the values of the corrosion rate and the corrosion width of the prestressed tendon in the detection and analysis, and improve the accuracy of quantitative identification of the corrosion damage of the prestressed tendon.

[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the purpose and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A concrete bridge prestressed tendon damage detection system, characterized by: It includes a magnetic detection unit, a signal processing unit and a damage detection unit; The magnetic detection unit is used to collect leakage magnetic field signals of the area to be detected; The signal processing unit is used to perform data processing on the collected leakage magnetic field signal to obtain processed data; The damage detection unit is used to extract features from the processed data to obtain feature parameters, input the feature parameters into a neural network model to train the network model, and output damage detection results using the trained network model.

2. The concrete bridge prestressed tendon damage detection system according to claim 1 is characterized in that: The magnetic detection unit includes an excitation device, a magnetic sensor and an auxiliary motion device; The excitation device uses a permanent magnet as a magnetic source, and a magnetic circuit is formed by the permanent magnet, the armature and the prestressed tendons; The magnetic sensor is used to collect the leakage magnetic field generated at the damaged part of the prestressed tendon; The auxiliary motion device is used to achieve accurate positioning and speed control of any three-dimensional spatial position, and drives the excitation device and the magnetic sensor to detect the designated area to be detected according to the planned path.

3. The concrete bridge prestressed tendon damage detection system according to claim 2 is characterized in that: The magnetic sensors are array-type; the array design arranges the magnetic sensors into two rows in the X and Y directions, with n magnetic sensors in each row, responsible for detecting the X and Z magnetic field components; wherein the moving direction along the beam is the X direction, the direction perpendicular to the X direction in the horizontal plane where the beam is located is the Y direction, and the direction perpendicular to the plane where the X and Y directions are located is the Z direction.

4. The concrete bridge prestressed tendon damage detection system according to claim 2 is characterized in that: Select 2 NdFeB permanent magnets of grade N35 or above, and the size of the magnets is determined according to the excitation requirements; select Q235 material to be processed into a rectangular armature of corresponding size, and set a magnet at each end of the armature to form a "U" shape structure with the two magnets to guide the magnetic lines of force and make the magnetic lines of force form a closed magnetic circuit.

5. The concrete bridge prestressed tendon damage detection system according to claim 1 is characterized in that: The collected leakage magnetic field signal is processed, including: The DC component and some noise of the leakage magnetic field signal are removed, and the weak voltage signal is amplified by operation.

6. The concrete bridge prestressed tendon damage detection system according to claim 5 is characterized in that: Remove the DC component of the leakage magnetic field signal, including: Eliminate the baseline in the magnetic flux leakage signal according to the following formula: Where S(i) is the magnetic flux leakage signal after baseline elimination; x i is the current data; n is the total number of collected data; Let the average value of the peak-to-valley value be The peak-to-valley value V obtained from the ath test a As the benchmark, S(i) is corrected, and the corrected data is S′(i): Where m is the total number of tests.

7. The concrete bridge prestressed tendon damage detection system according to claim 6 is characterized in that: Extract the peak-to-valley value of the magnetic flux leakage signal according to the following steps: S1. Set the peak search flag to flag = 1. When it is 1, search for peaks in the test data. When it is 0, search for valleys in the test data. S2. Assume that the initial value of the peak value Mx is -inf, the initial value of the valley value Mn is inf, and the peak-valley value threshold is delta; S3. The current value of the detection data is Di, and it is determined whether the current value Di is greater than the peak value Mx. If it is true, Mx = Di, otherwise, Mx remains unchanged; S4. Determine whether the current value Di is less than the valley value Mn. If so, set Mn = Di, otherwise, Mn remains unchanged; S5. If flag is 1, determine whether Di is less than Mx-delta. If so, set Mn = Di. Otherwise, Mn remains unchanged and flag is changed to 0. S6. If flag is 0, determine whether Di is greater than Mn+delta. If so, set Mx=Di. Otherwise, keep Mx unchanged and change flag to 1. S7. Determine whether the current data is the last data. If not, obtain the next data and repeat S3 to S6 until all data are detected and determined.

8. The concrete bridge prestressed tendon damage detection system according to claim 1 is characterized by: The characteristic parameters include the peak-to-valley difference of the axial signal, the peak span of the axial signal, the peak-to-valley difference of the normal signal, the peak-to-valley spacing of the normal signal, and the lift-off value of the magnetic sensor.

9. The concrete bridge prestressed tendon damage detection system according to claim 8, characterized in that: Input the feature parameters into the neural network model to train the network model, including: Select a suitable neural network model, set the activation function of the neural network to Sigmoid, the training error to 0.01, the learning rate to 0.01, the number of hidden layer nodes to 10, the number of input layer nodes to 5, and the number of output layer nodes to 2; Characteristic parameters under several working conditions are obtained to form sample data; 90% of the sample data are used as training samples, and the remaining 10% of the sample data are used as test samples.

Images