Fastener defect detection method and system based on ultrasonic CT

Abstract

The invention provides a fastener defect detection method and system based on ultrasonic CT, and belongs to the field of acoustic detection.The method comprises the steps that an acquisition device is arranged; the rotating lifting platform acts to enable the fastener to rotate by a set rotating angle, ultrasonic transmitters of the transmitting array sequentially transmit ultrasonic signals, the ultrasonic signals are received by all ultrasonic receivers of the receiving array to form training defect signals, and the training defect signals are used for generating corresponding training slice images in an FWI mode; all the training slice images and the training defect signals corresponding to the training slice images form training data pairs, and the neural network is trained until the neural network converges; the acquisition device acquires a set number of defect signals of the fastener, each defect signal is input into the convergent neural network to obtain a slice image corresponding to the defect signal, and three-dimensional data which can be checked by a user is formed. The method can effectively improve the detection precision and efficiency of the internal defect of the fastener.

Description

Technical Field

[0001] The present invention belongs to the field of acoustic detection, and particularly relates to a method and system for detecting fastener defects based on ultrasonic CT. Background Art

[0002] The quality of fasteners (such as bolts and nuts) is directly related to the stability and safety of mechanical equipment. In traditional fastener quality inspection, it mainly relies on manual visual inspection, that is, inspectors observe the surface of fasteners with the naked eye in a high-brightness environment to identify potential defects. However, this method is limited by the resolution of the human eye and the subjective judgment of inspectors, and it is difficult to ensure the accuracy and consistency of the inspection results. Especially in large-scale production, the efficiency and accuracy of manual inspection are difficult to meet market demands.

[0003] Deep learning has greatly promoted the development and application of machine vision technology. Through high-resolution cameras and advanced image processing algorithms, it can automatically identify and analyze defects on the surface of fasteners, such as scratches, depressions, dimensional deviations, etc., significantly improving the efficiency and accuracy of inspection and reducing the interference of human factors. However, conventional visual imaging technology is difficult to observe internal pores, cracks and other defects. Therefore, existing machine vision technology has limitations in detecting internal defects of fasteners. If fasteners with internal defects are used in key mechanical equipment, it may lead to serious safety accidents, resulting in huge economic losses and even casualties. Especially in key fields such as aerospace, industrial ships, mechatronics and transportation, the quality and safety requirements for fasteners are extremely high, and the demand for detecting internal defects of fasteners in these fields is particularly urgent. Summary of the Invention

[0004] The purpose of the present invention is to propose a method and system for detecting fastener defects based on ultrasonic CT, which can effectively improve the accuracy and efficiency of detecting internal defects of fasteners.

[0005] The present invention is achieved through the following technical solutions:

[0006] A method for detecting fastener defects based on ultrasonic CT includes the following steps:

[0007] Step S1, set up a collection device. The collection device includes a rotary lifting table for placing the fastener and making the fastener lift and rotate, and a transmitting array and a receiving array that are relatively spaced on both sides of the rotary lifting table. The transmitting array includes a plurality of ultrasonic transmitters, and the receiving array includes a plurality of ultrasonic receivers. The rotary lifting table operates to make the fastener lift to be within the scanning range of each ultrasonic transmitter and each ultrasonic receiver;

[0008] Step S2: If the number of currently acquired training slice images reaches the set number, proceed to step S3; otherwise, rotate the lifting platform to make the fastener rotate by a set rotation angle. Each ultrasonic transmitter in the transmitting array sequentially transmits ultrasonic signals. Each ultrasonic signal is received by all ultrasonic receivers in the receiving array after passing through the fastener, forming a training defect signal in the form of a complex matrix. The training defect signal is used in the FWI method to generate the corresponding training slice image, and repeat step S2;

[0009] Step S3: Form training data pairs from all training slice images and the training defect signals corresponding to each training slice image, and use the dataset composed of the training data pairs to train the neural network until the neural network converges;

[0010] Step S4: The acquisition device acquires a set number of defect signals of the fastener. Each defect signal is input into the converged neural network to obtain the slice image corresponding to the defect signal, and each slice image is combined to form three-dimensional data for the user to view.

[0011] Further, in step S1, the acquisition device further includes a medium pool, a medium disposed in the medium pool, a driving mechanism, and two ultrasonic brackets. The fastener, the transmitting array, and the receiving array are all located in the medium. A part of the rotating lifting platform is located in the medium pool and the other part extends downward out of the medium pool. The driving mechanism is disposed outside the medium pool and is connected to the rotating lifting platform. The two ultrasonic brackets are oppositely disposed in the medium pool. The transmitting array and the receiving array are respectively disposed on the two ultrasonic brackets so as to be able to move away from or close to the fastener.

[0012] Further, in step S1, the rotating lifting platform includes a lead screw nut pair and a placement plate disposed at the upper end of the lead screw nut pair. The driving mechanism includes a stepping motor connected to the lower end of the lead screw nut pair. The fastener is placed on the placement plate and the axis of the fastener is aligned with the axis of the lead screw nut pair.

[0013] Further, in step S1, the transmitting array and the receiving array can both be in a horizontal position or a vertical position.

[0014] Further, in step S1, the acquisition device further includes two phased array drivers respectively connected to the transmitting array and the receiving array. In step S2, the spatial position coordinates of each ultrasonic transmitter and each ultrasonic receiver are respectively obtained through the two phased array drivers, so that the signals received by the receiving array are represented in the form of a complex matrix.

[0015] Further, in step S2, when training the inversion neural network, the objective function is to minimize the L2 norm difference between the actually observed waveform data and the waveform data obtained through the inversion neural network.

[0016] The present invention is also achieved through the following technical solutions:

[0017] A fastener defect detection system based on ultrasonic CT includes a control module and an acquisition device connected to the control module. The acquisition device includes a rotary lifting table for placing the fastener and enabling the fastener to lift and rotate, and a transmitting array and a receiving array that are relatively spaced apart on both sides of the rotary lifting table. The transmitting array includes a plurality of ultrasonic transmitters, and the receiving array includes a plurality of ultrasonic receivers. When starting to acquire data, the control module controls the rotary lifting table to move so that the fastener lifts to be within the scanning range of each ultrasonic transmitter and each ultrasonic receiver. If the number of training slice images currently obtained does not reach the set number, the control module controls the rotary lifting table to move so that the fastener rotates by a set rotation angle, and controls each ultrasonic transmitter in the transmitting array to sequentially transmit ultrasonic signals. Each ultrasonic signal passes through the fastener and is received by all the ultrasonic receivers in the receiving array, forming a training defect signal in the form of a complex matrix. The control module generates corresponding training slice images from the training defect signal by means of FWI. If the number of training slice images currently obtained reaches the set number, the control module forms training data pairs from all the training slice images and the training defect signals corresponding to each training slice image, and uses the data set composed of the training data pairs to train a neural network until the neural network converges. During actual detection, the control module controls the acquisition device to acquire a set number of defect signals of the fastener, inputs each defect signal into the converged neural network to obtain the slice image corresponding to the defect signal, and combines each slice image to form three-dimensional data for the user to view.

[0018] Further, the acquisition device further includes a medium pool, a medium disposed in the medium pool, a driving mechanism, and two ultrasonic brackets. The fastener, the transmitting array, and the receiving array are all located in the medium. A part of the rotary lifting table is located in the medium pool and the other part extends downward out of the medium pool. The driving mechanism is disposed outside the medium pool and is connected to the rotary lifting table. The two ultrasonic brackets are oppositely arranged in the medium pool. The transmitting array and the receiving array are respectively arranged on the two ultrasonic brackets so as to be able to move away from or close to the fastener.

[0019] Further, the rotary lifting table includes a lead screw nut pair and a placement plate provided at the upper end of the lead screw nut pair. The driving mechanism includes a stepping motor connected to the lower end of the lead screw nut pair. The fastener is placed on the placement plate and the axis of the fastener is aligned with the axis of the lead screw nut pair.

[0020] Further, the control module is implemented by a computer.

[0021] The present invention has the following beneficial effects:

[0022] 1. The present invention uses a collection device to collect defect signals. In the training stage, the collected defect signals are input into a trained inversion neural network for inversion to obtain training slice images. When the number of training slice images reaches a set number, the defect signals and the corresponding training slice images form training data pairs, and a neural network is trained using the data set composed of the training data pairs until the neural network converges. In the detection stage, the defect signals collected by the collection device are input into the converged neural network to obtain slice images, and all the slice images are combined to form three-dimensional data for the user to view, thus completing defect detection. In this process, an end-to-end neural network reconstructs the defect signals collected by the collection device, which can effectively improve the detection efficiency and detection accuracy. The collection device includes a rotary lifting table for placing fasteners and lifting and rotating the fasteners, and a transmitting array and a receiving array relatively and spaced apart on both sides of the rotary lifting table. The transmitting array includes a plurality of ultrasonic transmitters, and the receiving array includes a plurality of ultrasonic receivers. During signal collection, the rotary lifting table moves so that the fastener is within the scanning range of each ultrasonic transmitter and each ultrasonic receiver. Each ultrasonic transmitter in the transmitting array sequentially emits ultrasonic signals, and all the ultrasonic receivers in the receiving array receive the ultrasonic signals after passing through the fastener, forming defect signals in the form of a complex matrix. The rotary lifting table rotates the fastener continuously at a set rotation angle to collect defect signals at multiple angles, so as to more accurately determine the location of internal defects in the fastener. The transmitting array and the receiving array respectively include a plurality of ultrasonic transmitters and a plurality of ultrasonic receivers, which can realize the simultaneous collection of multiple groups of acoustic wave signals with defects, thus solving the problem that the data required by the ultrasonic CT imaging method is proportional to the imaging accuracy. Compared with a single-element ultrasonic sensor, it also avoids the problem of frequent movement of the ultrasonic sensor position, thereby improving the imaging accuracy and the efficiency of data collection. During the collection process, there is no need for manual work such as changing wires and reading signals, thus saving labor costs and being more convenient.

[0023] 2. The transmitting array and the receiving array can be located in a horizontal position or a vertical position, and can be placed in a horizontal position or a vertical position according to needs, so as to obtain cross-sectional views from different perspectives, which is more conducive to the detection of internal defects in fasteners. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The present invention will be further described in detail below with reference to the accompanying drawings.

[0025] Figure 1 It is a schematic structural diagram of the collection device of the present invention.

[0026] Figure 2 It is a flowchart of the present invention.

[0027] Figure 3 It is a schematic diagram of the detection principle of the present invention.

[0028] Figure 4 Schematic diagram of the scanning strategy of the present invention.

[0029] Figure 5 Schematic diagram of obtaining the slice image of the present invention.

[0030] Wherein, 1, fastener; 2, medium pool; 21, medium; 3, rotary lifting table; 4, driving mechanism; 5, transmitting array; 6, receiving array; 7, ultrasonic bracket; 8, moving mechanism. Detailed implementation manners

[0031] The fastener defect detection system based on ultrasonic CT includes a control module and a collection device connected to the control module. As Figure 1 shown, the collection device includes a medium pool 2, a medium 21 arranged in the medium pool 2, a rotary lifting table 3 for placing the fastener 1 and lifting and rotating the fastener 1, a driving mechanism 4 for driving the rotary lifting table 3 to act, a transmitting array 5 and a receiving array 6 relatively and spaced apart on both sides of the rotary lifting table 3, two ultrasonic brackets 7, two moving mechanisms 8, and two phased array drivers respectively connected to the transmitting array 5 and the receiving array 6. The transmitting array 5 includes 64 ultrasonic transmitters, and the receiving array 6 includes 64 ultrasonic receivers. In this embodiment, the medium 21 is water, and the fastener 1, the transmitting array 5 and the receiving array 6 are all located in the medium 21. In other embodiments, the medium 21 can also be gel.

[0032] The rotary lifting table 3 includes a lead screw nut pair and a placement tray arranged at the upper end of the lead screw nut pair. The driving mechanism 4 includes a stepping motor connected to the lower end of the lead screw nut pair. The stepping motor is connected to the control module to be controlled by the control module to act. The fastener 1 is placed on the placement tray, and the axis of the fastener 1 is aligned with the axis of the lead screw nut pair. A part of the lead screw nut pair is located in the medium pool 2, and the other part passes downward through the medium pool 2. The driving mechanism 4 is arranged outside the medium pool 2 and connected to the rotary lifting table 3. The two ultrasonic brackets 7 are relatively arranged in the medium pool 2. The transmitting array 5 and the receiving array 6 are respectively arranged on the two ultrasonic brackets 7 through the two moving mechanisms 8 and can be far away from or close to the fastener 1. Among them, the connection structure between the stepping motor and the lead screw nut pair is the prior art. The placement tray is provided with a positioning mechanism for aligning the axis of the fastener 1 with the axis of the lead screw nut pair, and the specific structure of this positioning mechanism is the prior art.

[0033] The ultrasonic bracket 7 includes a support rod vertically arranged in the medium pool 2. The moving mechanism 8 is arranged on the support rod through an existing slider mechanism and can move up and down. The transmitting array 5 is arranged on the moving mechanism 8. In this embodiment, the moving mechanism 8 is an existing wireless module, and the moving mechanism 8 enables the transmitting array 5 to move horizontally.

[0034] In this embodiment, the transmitting array 5 and the receiving array 6 can be located in the horizontal position or the vertical position. When the 64 ultrasonic transmitters are arranged in a 64×1 pattern, it is in the horizontal position, and when the 64 ultrasonic transmitters are arranged in a 1×64 pattern, it is in the vertical position. This can be achieved through an existing rotating clamping mechanism connected between the transmitting array 5 and the moving mechanism 8. The receiving array 6 is the same as the transmitting array 5.

[0035] In this embodiment, the control module is implemented by a computer.

[0036] As Figures 2 to 4 shown, the fastener defect detection method based on ultrasonic CT includes the following steps:

[0037] Step S1: Set up the acquisition device. The acquisition device includes a rotary lifting table 3 for placing the fastener 1 and enabling the fastener 1 to lift and rotate, and a transmitting array 5 and a receiving array 6 that are relatively spaced apart on both sides of the rotary lifting table 3. The transmitting array 5 includes a plurality of ultrasonic transmitters, and the receiving array 6 includes a plurality of ultrasonic receivers. When starting the acquisition, the rotary lifting table 3 operates to lift the fastener 1 so that it is within the scanning range of each ultrasonic transmitter and each ultrasonic receiver.

[0038] Step S2: If the number of currently acquired training slice images reaches the set number, enter step S3; otherwise, the rotary lifting table 3 operates to rotate the fastener 1 by a set rotation angle. Each ultrasonic transmitter of the transmitting array 5 sequentially emits ultrasonic signals, and all ultrasonic receivers of the receiving array 6 receive the ultrasonic signals after passing through the fastener 1, forming a training defect signal in the form of a complex matrix. The training defect signal is used to generate a corresponding training slice image by the FWI method, and step S2 is repeated.

[0039] In this embodiment, the control module controls the rotation of the stepper motor through pulses. Each pulse interval is 1 s, and the stepper motor rotates by a step size of 2° each time (i.e., the set rotation angle is 2°). The stepper motor completes one full rotation after 180 pulses. At this time, the screw-nut pair drives the fastener 1 to move vertically by 30 mm. During each pulse interval, the transmitting array 5 and the receiving array 6 complete the transmission and reception of ultrasonic waves.

[0040] The control module obtains the spatial position coordinates of each ultrasonic transmitter and each ultrasonic receiver through two phased array drivers respectively, so that the signals received by the receiving array 6 are represented in the form of a complex matrix. More specifically, the control module controls the 64 ultrasonic transmitters of the transmitting array 5 to sequentially transmit ultrasonic waves through the phased array driver. The ultrasonic waves transmitted by each ultrasonic transmitter are received by all ultrasonic receivers, and finally a complex matrix with a size of 64×64 is obtained. After being filtered, this complex matrix is input into the inversion neural network for inversion. When data is collected, the fastener 1 rotates and rises at the same time. The inverted image is a cross-sectional view of the acquisition position at each step, that is, the obtained image is a slice with rotation. Therefore, it is also necessary to obtain the inverse solution in the reverse direction of the rotation angle of the stepping motor for the 180 frames of images obtained by rotating the stepping motor one week according to the rotation angle of the stepping motor, so as to obtain the slice image of the fastener 1. In order to enable the slice image to reach the required resolution, the slice images can be summarized and interpolated.

[0041] When training the inversion neural network, the L2 norm difference between the actually observed waveform data and the waveform data obtained through the inversion neural network is minimized as the objective function, that is, the loss function adopted in the training of the inversion neural network is expressed as where represents the wave field data received by the ultrasonic receiver, represents the reconstructed wave field data. The propagation space is divided into layered propagation intervals, and the wave field of the entire space is iteratively propagated and fitted from the ultrasonic receiver upwards.

[0042] Step S3: Form training data pairs from all training slice images and the training defect signals corresponding to the training slice images, and use the data set composed of the training data pairs to train the neural network until the neural network converges;

[0043] Step S4: The acquisition device acquires a set number of defect signals of the fastener 1. Each defect signal is input into the converged neural network to obtain the slice image corresponding to the defect signal. The slice images are combined to form three-dimensional data for the user to view, as Figure 5 shown. In this embodiment, the neural network uses the UNet neural network.

[0044] The above is only a preferred embodiment of the present invention, and thus cannot limit the scope of implementation of the present invention. That is, equivalent changes and modifications made according to the scope of the patent application of the present invention and the content of the specification should still fall within the scope covered by the patent of the present invention.

Claims

1. A fastener defect detection method based on ultrasonic CT, characterized in that: The steps include: Step S1, setting a collection device, the collection device comprising a rotating lifting platform for placing fasteners and causing the fasteners to be lifted and rotated, and a transmitting array and a receiving array arranged at opposite sides of the rotating lifting platform with a relative interval, the transmitting array comprising a plurality of ultrasonic transmitters, the receiving array comprising a plurality of ultrasonic receivers, the rotating lifting platform moves to cause the fasteners to be lifted and lowered so as to be located within the scanning range of each ultrasonic transmitter and each ultrasonic receiver; Step S2: If the number of currently acquired training slice images reaches the set number, go to step S3; otherwise, the rotating lifting platform causes the fastener to rotate by the set rotation angle, and each ultrasonic transmitter of the transmitting array transmits an ultrasonic signal in turn. Each ultrasonic signal is received by all ultrasonic receivers of the receiving array after passing through the fastener, forming a training defect signal in the form of a complex matrix. The training defect signal is used through FWI to generate a corresponding training slice image, and step S2 is repeated; Step S3, forming a training data pair from all training slice images and the training defect signals corresponding to each training slice image, and training the neural network using the data set consisting of the training data pairs until the neural network converges; Step S4: The acquisition device acquires a set number of defect signals of the fasteners, and each defect signal is input into a converged neural network to obtain a slice image corresponding to the defect signal, and each slice image is combined to form three-dimensional data that can be viewed by the user.

2. The fastener defect detection method based on ultrasonic CT according to claim 1, characterized in that: In step S1, the collection device also includes a medium pool, a medium arranged in the medium pool, a driving mechanism, and two ultrasonic brackets. The fastener, the transmitting array, and the receiving array are all located in the medium. A portion of the rotating lifting platform is located in the medium pool, and another portion passes downward out of the medium pool. The driving mechanism is arranged outside the medium pool and connected to the rotating lifting platform. The two ultrasonic brackets are relatively arranged in the medium pool. The transmitting array and the receiving array can be respectively arranged on the two ultrasonic brackets away from or close to the fastener.

3. The fastener defect detection method based on ultrasonic CT according to claim 2 is characterized in that: In step S1, the rotary lifting platform includes a screw nut pair and a storage tray arranged at the upper end of the screw nut pair, the driving mechanism includes a stepper motor connected to the lower end of the screw nut pair, the fastener is placed on the storage tray and the axis of the fastener is aligned with the axis of the screw nut pair.

4. A fastener defect detection method based on ultrasonic CT according to claim 1, 2 or 3, characterized in that: In step S1, the transmitting array and the receiving array may be located in a horizontal position or a vertical position.

5. A fastener defect detection method based on ultrasonic CT according to claim 1, 2 or 3, characterized in that: In step S1, the acquisition device also includes two phased array drivers respectively connected to the transmitting array and the receiving array. In step S2, the spatial position coordinates of each ultrasonic transmitter and each ultrasonic receiver are respectively obtained by the two phased array drivers so that the signal received by the receiving array is represented in the form of a complex matrix.

6. A fastener defect detection method based on ultrasonic CT according to claim 1, 2 or 3, characterized in that: In step S2, when the inversion neural network is trained, the objective function is to minimize the L2 norm difference between the waveform data actually observed and the waveform data obtained by the inversion neural network.

7. A fastener defect detection system based on ultrasonic CT, characterized in that: The invention comprises a control module and a collection device connected to the control module, wherein the collection device comprises a rotating lifting platform for placing fasteners and making the fasteners rise and fall and rotate, and a transmitting array and a receiving array arranged at opposite sides of the rotating lifting platform at a relative interval, wherein the transmitting array comprises a plurality of ultrasonic transmitters, and the receiving array comprises a plurality of ultrasonic receivers. When the collection starts, the control module controls the rotating lifting platform to move the fasteners to rise and fall so as to be located within the scanning range of each ultrasonic transmitter and each ultrasonic receiver. If the number of training slice images currently acquired does not reach the set number, the control module controls the rotating lifting platform to move the fasteners to rotate the set rotation angle, and controls each ultrasonic transmitter of the transmitting array to transmit ultrasonic signals in sequence, and each ultrasonic signal is received by the receiving array after passing through the fasteners. All ultrasonic receivers receive the training defect signals in the form of complex matrices. The control module generates corresponding training slice images through FWI. If the number of currently acquired training slice images reaches the set number, the control module forms training data pairs with all the training slice images and the training defect signals corresponding to each training slice image. The neural network is trained using the data set composed of the training data pairs until the neural network converges. During actual detection, the control module controls the acquisition device to acquire a set number of defect signals of the fasteners, inputs each defect signal into the converged neural network to obtain the slice image corresponding to the defect signal, and combines each slice image to form three-dimensional data that can be viewed by the user.

8. The fastener defect detection system based on ultrasonic CT according to claim 7, characterized in that: The collection device also includes a medium pool, a medium arranged in the medium pool, a driving mechanism, and two ultrasonic brackets. The fastener, the transmitting array, and the receiving array are all located in the medium. A portion of the rotating lifting platform is located in the medium pool, and another portion passes downward out of the medium pool. The driving mechanism is arranged outside the medium pool and connected to the rotating lifting platform. The two ultrasonic brackets are relatively arranged in the medium pool. The transmitting array and the receiving array can be respectively arranged on the two ultrasonic brackets away from or close to the fastener.

9. The fastener defect detection system based on ultrasonic CT according to claim 8, characterized in that: The rotary lifting platform includes a screw nut pair and a storage tray arranged at the upper end of the screw nut pair, the driving mechanism includes a stepping motor connected to the lower end of the screw nut pair, the fastener is placed on the storage tray and the axis of the fastener is aligned with the axis of the screw nut pair.

10. A fastener defect detection system based on ultrasonic CT according to claim 7, 8 or 9, characterized in that: The control module is implemented by a computer.

Images