3D Printing Electromagnetic Wave Absorbing Concrete Device Based on Computer Vision

Abstract

The present invention is a 3D printing electromagnetic wave absorbing concrete device based on computer vision, which includes a control unit, a resin dry material storage unit, a sorting and feeding unit, and a water absorption unit; the water absorption unit is connected to a weighing device and the resin dry material storage unit; the lower feed port of the sorting and feeding unit is connected to the lower solid discharge port of the water absorption unit. In the sorting and feeding unit, the wet water-absorbing resin is sorted directionally along the spiral screening and transportation groove, and the wet water-absorbing resin is sent to the outlet of the sorting and feeding unit, and then connected to a linear feeding track; an opening for inserting a metal tube is provided on the side wall of the detachable printing nozzle. The end of the linear feeding track is fixed near the opening of the conical nozzle of the detachable printing nozzle through a connecting piece. One end of the metal tube is connected to the end of the linear feeding track, and the other end extends into the detachable printing nozzle. This device can achieve the fusion of concrete and water-absorbing resin, and multiple water-absorbing resins and concrete are extruded together through the print head, so that the concrete forms a porous structure.

Description

Technical Field

[0001] The present invention belongs to the technical field of electromagnetic wave absorbing concrete preparation devices, and particularly relates to a 3D printing electromagnetic wave absorbing concrete device based on computer vision. Background Art

[0002] The absorption of electromagnetic waves by porous materials is mainly due to the reflection, scattering, and interference of electromagnetic waves by their porous structures, resulting in the attenuation of electromagnetic waves. Porous wave-absorbing materials are mainly closed microporous materials, such as porous ceramics, porous metals, EPS-filled porous cement-based materials, etc. Porous concrete has interconnected open pore structures and relatively large pore diameters, allowing sound waves to enter and propagate inside. Due to the viscosity of air and the inherent damping characteristics of the material, the incoming sound energy is continuously dissipated, thus playing a sound absorption role. In recent years, 3D printing technology has become a social hotspot, and its printing materials include plastics, metals, ceramics, soft materials, concrete, etc. 3D printing concrete technology has many advantages such as free design and construction, high automation, fast construction speed, low labor cost, and low environmental pollution, and has received extensive attention and significant development in the field of civil engineering in recent years. 3D printed bridges, 3D printed houses, etc. can be seen everywhere, which largely confirms the feasibility of applying 3D printing technology to the field of civil engineering. When preparing porous electromagnetic wave absorbing concrete with a spherical cavity structure using spherical water-absorbing resin, if the spherical water-absorbing resin and concrete are put into a mixer and stirred together, its original structure will be damaged, making it unable to achieve the expected effect.

[0003] To prepare 3D printing electromagnetic wave absorbing concrete materials with a spherical water-absorbing resin as the porous structure, it is necessary to arrange the water-absorbing resin on its surface while printing the concrete. The running path of the 3D printing concrete print head is variable, so a device is needed that can extrude the water-absorbing resin together with the concrete without damage when the concrete is extruded in a layer, so as to adapt to the free construction process of 3D printing technology. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the technical problem to be solved by the present invention is to provide a 3D printing electromagnetic wave absorbing concrete device based on computer vision. This device can achieve the fusion of concrete and water-absorbing resin, and multiple water-absorbing resins and concrete are extruded together through the print head, making the concrete form a porous structure. The porous structure can improve the matching of its surface impedance and free space impedance, making it easier for electromagnetic waves to enter the material interior, thereby increasing its loss attenuation; the incident electromagnetic waves will rapidly attenuate after multiple reflections and scatterings in the porous structure, so that the porous concrete has good electromagnetic wave absorption performance.

[0005] The technical solution adopted by the present invention to solve the above technical problem is:

[0006] In a first aspect, the present invention provides a 3D printing electromagnetic wave absorbing concrete device based on computer vision, which includes a spiral extrusion unit, a detachable printing nozzle, a water weighing device and a control unit. The device further includes a dry resin storage unit, a sorting and feeding unit, and a water absorption unit.

[0007] The sorting and feeding unit and the spiral extrusion unit are connected by a fixing plate. A linear module is arranged at the lower part of the fixing plate, and the fixing plate can drive the sorting and feeding unit and the spiral extrusion unit to move back and forth on the linear module.

[0008] The dry resin storage unit is used to store dry water-absorbing resin.

[0009] The water weighing device is used to control the amount of water added.

[0010] The water absorption unit is connected to the water guide pipe of the water weighing device and the feeding port of the dry resin storage unit. Inside the water absorption unit, the dry water-absorbing resin forms wet water-absorbing resin after being saturated with water.

[0011] The lower end feeding port of the sorting and feeding unit is connected to the lower solid discharge port of the water absorption unit. The sorting and feeding unit has a spiral screening and transportation groove. Inside the sorting and feeding unit, the wet water-absorbing resin is sorted directionally along the spiral screening and transportation groove, and the wet water-absorbing resin is sent to the outlet of the sorting and feeding unit. The outlet of the sorting and feeding unit is connected to a linear feeding track.

[0012] The side wall of the detachable printing nozzle is provided with an opening for inserting a metal tube. The end of the linear feeding track is fixed near the opening of the conical nozzle of the detachable printing nozzle through a connecting piece. One end of the metal tube is connected to the end of the linear feeding track, and the other end extends into the detachable printing nozzle and is coaxial with the nozzle of the detachable printing nozzle. There is a gap between the metal tube and the nozzle of the detachable printing nozzle.

[0013] A CCD camera is arranged at the outlet of the sorting and feeding unit for photographing and identifying the materials at the outlet of the sorting and feeding unit. A waste material falling port is arranged on the linear feeding track near the outlet of the sorting and feeding unit. The opening and closing of the waste material falling port is controlled by an electric valve. When it is identified that the material at the outlet of the sorting and feeding unit is not a complete sphere, the electric valve is controlled to open, so that the material falls into the waste material falling port. If it is identified that the material is a complete sphere, the electric valve is not opened, and the material rolls into the metal tube through the linear feeding track.

[0014] A waste bucket is arranged below the waste material falling port.

[0015] The water weighing device, the dry resin storage unit, the water absorption unit, the electric valve, the sorting and feeding unit, and the CCD camera are all electrically connected to the control unit.

[0016] The target recognition algorithm is loaded in the control unit, which can quickly recognize the shape of the wet water-absorbing resin. If it is recognized as a complete spherical wet water-absorbing resin, it will be directly sent to the metal tube through the linear feeding track. If it is recognized as an irregular or broken wet water-absorbing resin, the electric valve at the waste material dropping port will be opened to let the material fall and enter the waste bucket.

[0017] The target recognition algorithm can quickly recognize the shape of the material within 0.3s - 0.5s. The materials reach the outlet position of the sorting and feeding unit one by one for individual recognition. After the previous material leaves the waste material dropping port, the control unit closes the electric valve and then determines whether to open the electric valve again according to the recognition result of the next material. The interval between two adjacent materials is adjusted according to the adjacent resin printing gap required in actual 3D printing, ensuring that only the complete spherical materials enter the metal tube after passing through the judgment one by one.

[0018] The target recognition algorithm is an improved YOLOv7 deep neural network. The improved YOLOv7 deep neural network includes a backbone network Backbone and a Head network. The input image is input into the backbone network Backbone, and in the backbone network Backbone, it passes through CBS1, CBS2, CBS1, CBS2, ELAN, MP-1, ELAN, MP-1, ELAN, MP-1, ELAN in sequence and then outputs SPPCSPC that connects to the Head network. SPPCSPC passes through a CBS3 and Upsample to output result A; the output of the penultimate ELAN is simultaneously spliced with result A through a CBS3. The spliced result passes through the first ELAN-W, CBS3, and Upsample to output result B; the output of the second ELAN is spliced with result B through a CBS3. The spliced result passes through the second ELAN-W and is simultaneously connected to MP-2 and REP.

[0019] The output of MP-2 and the output of result A after being processed by the first ELAN-W are spliced. The spliced output is processed by the third ELAN-W, then spliced with the output of SPPCSPC through an MP-2, and finally processed by the fourth ELAN-W.

[0020] The outputs of the second, third, and fourth ELAN-W mentioned above are all connected to a CBM block through a REP. The outputs of the three CBM blocks are connected to each other to output the large-pixel feature map as the output image of the final network.

[0021] CBS1, CBS2, and CBS3 have the same structure, each including a convolutional layer, an LN layer, and a DSiLU function. For CBS1, k = 3, s = 1; for CBS2, k = 3, s = 2; for CBS3, k = 1, s = 1, where k is the convolutional kernel size and s is the stride.

[0022] REP consists of a convolutional layer and an LN layer connected in series to form one branch. There are two such identical branches connected in parallel and then connected in parallel with a branch formed by an LN layer. The outputs of the three branches are processed by addition to obtain the output of REP.

[0023] MP-1 and MP-2 have the same structure, each including two branches. One branch is formed by connecting a max pooling layer and CBS3, and the other branch is formed by connecting CBS3 and CBS2. The outputs of the two branches are concatenated to obtain the output of MP-1 or MP-2. Among them, the number of hidden layers of MP-1 is i / 2, and the number of hidden layers of MP-2 is i, where i represents the number of input nodes.

[0024] The structure of SPPCSPC is as follows: The data passes through the CBS3, CBS1, CBS3 modules once and then undergoes three max pooling processes. The results of the three max pooling processes are concatenated with the output of the last CBS3. After that, it passes through a CBS3 and is directly concatenated with the output of the input of SPPCSPC after being processed by a CBS3. Finally, it passes through a CBS3 to obtain the output of SPPCSPC.

[0025] The input of ELAN passes through a max pooling layer and the output result E of CBS3. The input of ELAN also sequentially passes through CBS3, CBS1, CBS1, CBS1 to obtain the output result F. Additionally, it passes through CBS3, CBS3 and CBS1, CBS3 and two CBS1 respectively to obtain results G, H, I. Then, E, F, G, H, I are jointly concatenated, and finally, the output of ELAN is obtained after passing through CBS3.

[0026] ELAN-W is similar in structure to ELAN. The difference is that there are four CBS1 on the result F branch. The input of ELAN sequentially passes through CBS3, CBS1, CBS1, CBS1, CBS1 to obtain the output result F. It also passes through CBS3, CBS3 and CBS1, CBS3 and two CBS1, CBS3 and three CBS1 respectively to obtain results G, H, I, J. Then, E, F, G, H, I, J are jointly concatenated, and finally, the output of ELAN is obtained after passing through CBS3.

[0027] The CBM block consists of a convolutional layer, an LN layer, and a sigmod function.

[0028] The end of the metal pipe is lower than the nozzle end of the detachable printing nozzle by a distance equal to the diameter of a spherical water-absorbing resin, so that the concrete slurry wraps the spherical water-absorbing resin before exiting the nozzle of the detachable printing nozzle and prints them together; or the end of the metal pipe is located 1-2 mm above the nozzle end of the detachable printing nozzle.

[0029] An electromagnetic valve 11 is provided at the connection position between the linear feeding track and the metal pipe to control whether to lay the resin material for the concrete. If no laying is carried out, the equipment for controlling the resin delivery stops working.

[0030] The particle size range of the dry water-absorbing resin is 1-2 mm. The dry water-absorbing resin is first saturated with water and then printed together with the concrete at the nozzle position. The water-absorbing resin is fed into the detachable printing nozzle of the spiral extrusion unit and extruded together with the concrete. The outlet of the metal pipe faces the axis of the detachable printing nozzle.

[0031] A grating counter is provided below the waste material falling inlet, and the grating counter is connected to the control unit.

[0032] In a second aspect, the present invention also protects a 3D printed electromagnetic wave absorbing concrete composite material. The composite material is composed of multiple repeating units as a whole. Each repeating unit is composed of 3D printed concrete and regular spherical wet water-absorbing resin wrapped inside the 3D printed concrete. A certain number of spherical wet water-absorbing resins are arranged in sequence on the concrete. After a certain number of spherical wet water-absorbing resins are cured and dried in the concrete, a regular cavity structure can be formed inside the concrete material.

[0033] Compared with the prior art, the beneficial effects of the present invention are:

[0034] The 3D printed concrete adopts a layer-by-layer stacking construction process. Existing printing equipment cannot implant spherical wet water-absorbing resins during the printing process. The device of the present invention is based on 3D printing technology and can orderly arrange a certain number of spherical wet water-absorbing resins on the concrete surface at the same time, which can improve the degree of freedom of implanting spherical wet water-absorbing resins in the 3D printed concrete structure. The 3D printed concrete electromagnetic wave absorbing concrete constructed has good wave absorbing ability and realizes construction freedom. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a schematic diagram of the overall structure of the 3D printed electromagnetic wave absorbing concrete device based on computer vision of the present invention.

[0036] Figure 2 It is a schematic diagram of the partial sectional structure of the 3D printed electromagnetic wave absorbing concrete device based on computer vision of the present invention.

[0037] Figure 3Schematic diagram of the improved YOLOv7 deep neural network of the present invention.

[0038] In the figure, there are a barrel 1, a spiral extrusion unit 2, a detachable printing nozzle 3, a fixing plate 4, a metal pipe 5, a sorting and feeding unit 6, a water absorption unit 7, a linear module 8, a linear feeding track 9, an electromagnetic valve 11, a stepping motor 12, a spiral screening and transporting trough 13, a water weighing device 14, a resin dry material storage unit 15, a feeding port 16, and a water guide pipe 10. Detailed implementation manners

[0039] The present invention will be further explained below in conjunction with embodiments and the accompanying drawings, but this is not intended to limit the protection scope of the present application.

[0040] The 3D printing electromagnetic wave absorbing concrete device based on computer vision of the present invention (hereinafter referred to as the device, see Figure 1-2 ) includes a spiral extrusion unit 2, a detachable printing nozzle 3, a sorting and feeding unit 6, a resin dry material storage unit 15, a water weighing device 14, a water absorption unit 7 and a control unit.

[0041] The sorting and feeding unit 6 and the spiral extrusion unit 2 are connected through a fixing plate 4. A linear module 8 is arranged below the fixing plate 4, and the fixing plate can drive the sorting and feeding unit 6 and the spiral extrusion unit 2 to move back and forth on the linear module 8;

[0042] The resin dry material storage unit 15 is used to store a certain amount of dry water-absorbing resin. The average diameter of the dry water-absorbing resin is 1-2 mm. The discharge amount of the resin dry material storage unit can be controlled by the control unit to control the amount of dry material added;

[0043] The spiral extrusion unit 2 includes a barrel 1, a stepping motor 12 and a spiral auger. The output end of the stepping motor 12 is connected to the spiral auger. The spiral auger is installed in the barrel. The slurry is sent into the barrel, and the stepping motor drives the spiral auger to extrude the slurry.

[0044] The water weighing device is an intelligent water weighing device, which is controlled by the control unit to control the amount of water added;

[0045] The water absorption unit 7 is connected to the water guide pipe 10 of the water weighing device and the feeding port 16 of the resin dry material storage unit. In the water absorption unit, the dry water-absorbing resin forms wet water-absorbing resin after being saturated with water;

[0046] The lower feeding port of the sorting and feeding unit 6 is connected to the lower solid discharge port of the water absorption unit. There is a spiral screening and transporting trough 13 in the sorting and feeding unit 6. In the sorting and feeding unit, the wet water-absorbing resin is sorted directionally along the spiral screening and transporting trough, and the wet water-absorbing resin is sent to the outlet of the sorting and feeding unit; the outlet of the sorting and feeding unit is connected to the linear feeding track 9, and the wet water-absorbing resin can automatically enter the linear feeding track 9;

[0047] In addition, a CCD camera is provided at the outlet of the sorting and feeding unit for photographing and identifying the materials at the outlet of the sorting and feeding unit. A waste material dropping port is provided on the linear feeding track near the outlet position of the sorting and feeding unit. The opening and closing of the waste material dropping port is controlled by an electric valve. When it is identified at the outlet of the sorting and feeding unit that the material is not a complete sphere, the electric valve is controlled to open, so that the material falls into the waste material dropping port; if it is identified that the material is a complete sphere, the electric valve is not opened, and the material rolls into the metal pipe 5 along the linear feeding track;

[0048] The water weighing device, the resin dry material storage unit 15, the water absorption unit, the electric valve, the sorting and feeding unit 6, and the CCD camera are all electrically connected to the control unit. The control unit can control the amount of water added to the water absorption unit by the water weighing device, the amount of dry material added to the water absorption unit by the resin dry material, identify the shape of the material according to the photographed pictures of the CCD camera, and then decide whether to open the electric valve for waste material dropping;

[0049] The target recognition algorithm is loaded in the control unit. In this embodiment, the YOLOv7 neural network is used as the target recognition algorithm model, which can quickly recognize the shape of the wet water-absorbing resin. If it is recognized as a complete spherical wet water-absorbing resin, it is directly sent to the metal pipe 5 through the linear feeding track; if it is recognized as an irregular or broken wet water-absorbing resin, the electric valve at the waste material dropping port is opened to make the material fall and enter the waste bucket.

[0050] The target recognition algorithm can quickly recognize the shape of the material within 0.3s - 0.5s. The materials reach the outlet position of the sorting and feeding unit one by one for individual recognition. After the previous material leaves the waste material dropping port position, the control unit closes the electric valve and then judges whether to open the electric valve again according to the recognition result of the next material. The interval between adjacent two materials is adjusted according to the adjacent resin printing gap required in actual 3D printing, ensuring that each material enters the metal pipe only after passing through the judgment in turn. In this application, the adjacent resin printing gap is determined according to the design requirements of 3D printed concrete. After determining this gap, the extrusion printing speed and the resin feeding speed are matched, and the directional transmission speed of sorting and feeding is also adjusted according to the actual situation, so that the operation of judging whether to drop the material for each material in turn at the waste material dropping port is realized, and the phenomenon that multiple materials fall into the waste material dropping port at the same time will not occur.

[0051] Preferably, the electric valve is installed 10 cm away from the outlet of the sorting and feeding unit 6 on the linear feeding track 9. In the embodiment, 1 complete spherical material is dropped per second, and the moving speed of the print head is adjustable, with several options including 20 mm / s, 40 mm / s, 60 mm / s, 80 mm / s, and 100 mm / s. Correspondingly, the gaps can be 20 mm, 40 mm, 60 mm, 80 mm, and 100 mm.

[0052] At a frequency of dropping 1 complete spherical material per second, the problem of multiple materials falling simultaneously will not occur.

[0053] The ratio of water to resin dry material storage in the water weigher is 1:2.

[0054] In the present invention, the sorting and feeding unit 6 can be realized according to the prior art. The size of the spiral screening and transporting trough 13 is adapted to the diameter of the spherical wet water-absorbing resin, and each wet water-absorbing resin is continuously arranged at intervals on the spiral screening and transporting trough 13.

[0055] The spiral extrusion unit 2 is driven by the inlet motor 12. An opening for inserting a metal tube is provided on the side wall of the detachable print head. The end of the linear feeding track is fixed near the opening of the conical nozzle of the detachable print head 3 through a connecting piece. One end of the metal tube 5 is connected to the end of the linear feeding track, and the other end extends into the detachable print head 3 and is coaxial with the nozzle of the detachable print head. There is a gap between the metal tube and the nozzle of the detachable print head.

[0056] The linear feeding track is made of transparent fiberglass material with a diameter of 3 mm. The end near the detachable print head is softened by welding with a metal tube. The other end of the metal tube extends from the side wall of the detachable print head to the central axis position of the middle part of the detachable print head. The central axis of the metal tube outlet coincides with the central axis of the detachable print head outlet.

[0057] Preferably, the end of the metal tube is lower than the end of the nozzle of the detachable print head by a distance equal to the diameter of a spherical water-absorbing resin, so that the concrete slurry wraps the spherical water-absorbing resin before exiting the nozzle of the detachable print head for printing together, ensuring that there will be a cavity at the central position of the printed material, thus ensuring the homogeneity of the material; or the end of the metal tube is not lower than the end of the nozzle of the detachable print head, so that resin materials are arranged in an orderly manner at the center of the current layer of concrete every time a layer of concrete is printed; or the end of the metal tube is 1 - 2 mm above the end of the nozzle of the detachable print head; or the outlet of the metal tube faces the axis of the detachable print head.

[0058] An electromagnetic valve 11 is provided at the connection position between the linear feeding track and the metal tube, which is used to control whether to lay resin materials for the concrete. If no laying is carried out, the equipment for resin conveying is controlled to stop working.

[0059] The electromagnetic valve, stepper motor, and linear module 8 described above are all connected to the control unit, which respectively controls whether to lay resin materials, whether to extrude and print, etc.

[0060] According to the result identified by the target recognition algorithm, the present invention can adjust the interval of the materials. If materials that do not meet the requirements appear continuously for a period of time, the feeding speed of the sorting and feeding unit can be increased, and the opening and closing time of the electric valve can be adaptively adjusted, etc., to perform vision-based adjustment, so as to ensure that the resin interval after printing is within a suitable range.

[0061] The working process of the device of the present invention is as follows: Workers put dry water-absorbing resin into the resin dry material storage unit 15, and through the control of the control unit, the dry water-absorbing resin is sent through the feeding port 16 into the quantitative material to the water-absorbing unit 7. At the same time, the control unit controls the intelligent water weighing device to weigh the amount of water required for the dry water-absorbing resin to be saturated with water, and then sends the water into the water-absorbing unit 7 through the water guide pipe. After the dry water-absorbing resin is completely saturated with water in the water-absorbing unit, the lower solid discharge port of the water-absorbing unit is opened, and the wet water-absorbing resin is sent into the sorting and feeding unit 6. The spherical wet water-absorbing resin is sorted one by one along the track in a spiral upward direction from the bottom through the spiral screening and transportation trough of the sorting and feeding unit to the outlet of the sorting and feeding unit. At the outlet position, the shape of the wet water-absorbing resin is identified by using the target recognition algorithm in the control unit through image acquisition, and it is determined whether to retain the material. If it is a complete spherical wet water-absorbing resin, the material diameter is sent through the linear feeding track and the metal pipe to the central axis position of the detachable printing nozzle 3; then the concrete and the spherical wet water-absorbing resin are printed together by a 3D printer to obtain 3D printed electromagnetic wave absorbing concrete, and the concrete contains complete spherical wet water-absorbing resin, which can obtain 3D printed electromagnetic wave absorbing concrete with regular cavities after curing and drying.

[0062] The electromagnetic valve can control the conveying and interruption process of resin feeding. When it is necessary to convey spherical wet water-absorbing resin, the electromagnetic valve will automatically open, and vice versa, the conveying process will be automatically interrupted. After the interruption, the linear feeding track moves in the reverse direction to convey the material to the bottom of the sorting and feeding unit. When it is necessary to convey again, the sorting and feeding unit will sort and supply the material again. In actual construction, only when the spherical wet water-absorbing resin is complete and the material rheological property parameters of the 3D printing background are appropriate, the 3D printing background will resume printing. A trial print of 1 meter in length is set to adjust the running state of the initial equipment.

[0063] In the present invention, the particle size range of the dry water-absorbing resin is about 1 mm. The dry water-absorbing resin is first saturated with water and then extruded together with the concrete at the nozzle position. The water-absorbing resin is sent into the detachable printing nozzle of the spiral extrusion unit and extruded together with the concrete. This printing method can minimize the damage to the spherical wet water-absorbing resin and is conducive to the free layout of the cavity positions during the 3D printing process.

[0064] The target recognition algorithm is an improved YOLOv7 deep neural network. The improved YOLOv7 deep neural network includes a backbone network Backbone and a Head network. The input image is input into the backbone network Backbone, and in the backbone network Backbone, it passes through CBS1, CBS2, CBS1, CBS2, ELAN, MP-1, ELAN, MP-1, ELAN, MP-1, ELAN in sequence and then outputs SPPCSPC that connects to the Head network. SPPCSPC passes through a CBS3 and Upsample to output result A; the output of the penultimate ELAN is simultaneously concatenated (Concat) with result A through a CBS3, and the concatenated result passes through the first ELAN-W, CBS3, and Upsample to output result B; the output of the second ELAN passes through a CBS3 and is concatenated with result B, and the concatenated result passes through the second ELAN-W and is simultaneously connected to MP-2 and REP.

[0065] The output of MP-2 and the output of result A after being processed by the first ELAN-W are concatenated. The concatenated output is processed by the third ELAN-W, then concatenated with the output of SPPCSPC through an MP-2, and finally processed by the fourth ELAN-W.

[0066] The outputs of the above-mentioned second, third, and fourth ELAN-W are all connected to a CBM block through a REP, and the outputs of the three CBM blocks are connected to each other to output a large-pixel feature map as the output image of the final network.

[0067] The structures of CBS1, CBS2, and CBS3 are the same, and each includes a convolutional layer, an LN layer, and a DSiLU function. CBS1: k = 3, s = 1; CBS2: k = 3, s = 2; CBS3: k = 1, s = 1, where k is the convolutional kernel size and s is the stride.

[0068] REP is composed of a convolutional layer and an LN layer in series to form one branch. There are two such identical branches in parallel and then in parallel with a branch composed of an LN layer. The outputs of the three branches are processed by addition (add) to obtain the output of REP.

[0069] MP-1 and MP-2 have the same structure, and each includes two branches. One branch is composed of a max pooling layer and a CBS3 connection, and the other branch is composed of a CBS3 and a CBS2 connection. The outputs of the two branches are concatenated to obtain the output of MP-1 or MP-2. Among them, the number of hidden layers of MP-1 is i / 2, and the number of hidden layers of MP-2 is i, where i represents the number of input nodes.

[0070] The structure of SPPCSPC is as follows: The data passes through the CBS3, CBS1, and CBS3 modules once and then undergoes three max-pooling operations. The results of the three max-pooling operations are concatenated with the output of the last CBS3. After that, it passes through a CBS3 and is directly concatenated with the output of the input of SPPCSPC after being processed by a CBS3. Finally, the output of SPPCSPC is obtained after passing through a CBS3.

[0071] The input of ELAN passes through a max-pooling layer and the output result E of CBS3. The input of ELAN also passes through CBS3, CBS1, CBS1, CBS1 in sequence and the output result F. Then, it is combined with the results G, H, I obtained by passing through CBS3, CBS3 and CBS1, CBS3 and two CBS1 respectively. Furthermore, E, F, G, H, I are concatenated together, and finally the output of ELAN is obtained after passing through CBS3.

[0072] ELAN-W is similar in structure to ELAN. The difference is that there are four CBS1 on the result F branch. The input of ELAN passes through CBS3, CBS1, CBS1, CBS1, CBS1 in sequence and the output result F. Then, it is combined with the results G, H, I, J obtained by passing through CBS3, CBS3 and CBS1, CBS3 and two CBS1, CBS3 and three CBS1 respectively. Furthermore, E, F, G, H, I, J are concatenated together, and finally the output of ELAN is obtained after passing through CBS3.

[0073] The CBM block consists of a convolutional layer, an LN layer, and a sigmod function.

[0074] Neural network training stage:

[0075] 1) Install an industrial camera at the outlet position of the sorting and feeding unit to obtain the video of the wet water-absorbing resin (the resin will roll forward in the video, and images from various angles are obtained), and obtain the spherical wet water-absorbing resin images frame by frame from the specified position in the video.

[0076] 2) Label the obtained spherical wet water-absorbing resin images to form a dataset with annotation information. The dataset with annotation information is divided into a training set and a test set according to a ratio of 7:3, and the annotation information is ball or non-ball.

[0077] 3) For the dataset with annotation information, use the dataset with annotation information to train the YOLOv7 neural network. Iterate through all the images in the training set in sequence. After 300 epochs of training, when the training model converges, that is, when the training gradient of the model is close to 0, stop training and extract the optimal network parameters for later implementation of prediction.

[0078] Implementation stage:

[0079] 1) Receive an image for object detection;

[0080] 2) If the image is recognized by the YOLOv7 neural network as a complete spherical wet water-absorbing resin image, the diameter is fed into a metal tube through a linear feeding track;

[0081] 3) If it is recognized as an irregular or broken wet water-absorbing resin, open the electric valve to fall into the waste bin through the waste inlet. A grating counter is set below the position of the waste inlet to determine the closing of the electric valve after the material has fallen, or the control unit sets the opening and closing time of the electric valve. In this embodiment, after the electric valve is opened, it can be closed before the shape of one material is judged;

[0082] 4) Until the printing is completed.

[0083] The present invention is mainly applied to the 3D printing construction of electromagnetic wave-absorbing concrete containing resin materials, and can automatically obtain the image of spherical wet water-absorbing resin during the 3D printing process and judge whether it is a complete sphere.

[0084] The 3D printed electromagnetic wave-absorbing concrete prepared by the present invention is composed of multiple repeating units to form a whole. Each layer of repeating unit is composed of concrete and a certain number of spherical wet water-absorbing resins. The certain number of spherical wet water-absorbing resins are arranged in sequence on the concrete. After the certain number of spherical wet water-absorbing resins are cured and dried in the concrete, a regular cavity structure can be formed inside the concrete material to improve the wave-absorbing effect.

[0085] The device of the present invention is based on 3D printing technology and can print spherical wet water-absorbing resins and concrete at the same time, which can enhance the wave-absorbing ability of the material. The construction process is simple and the cost is not high, with good engineering feasibility and application value, which will greatly promote the practical application of 3D printing technology in the civil engineering industry.

[0086] Matters not described in the present invention apply to the prior art.

Claims

1. A 3D printing electromagnetic wave absorbing concrete device based on computer vision, comprising a spiral extrusion unit, a detachable printing nozzle, a water weigher and a control unit, characterized in that, The device further includes a dry resin storage unit, a sorting and feeding unit, and a water absorption unit. The sorting and feeding unit and the spiral extrusion unit are connected by a fixing plate. A linear module is arranged at the lower part of the fixing plate, and the fixing plate can drive the sorting and feeding unit and the spiral extrusion unit to move back and forth on the linear module. The dry resin storage unit is used to store dry water-absorbing resin. The water weigher is used to control the amount of water added. The water absorption unit is connected to the water guide pipe of the water weigher and the feeding port of the dry resin storage unit. Inside the water absorption unit, the dry water-absorbing resin forms wet water-absorbing resin after being saturated with water. The lower feeding port of the sorting and feeding unit is connected to the lower solid discharging port of the water absorption unit. There is a spiral screening and transporting groove in the sorting and feeding unit. Inside the sorting and feeding unit, the wet water-absorbing resin is sorted directionally along the spiral screening and transporting groove, and the wet water-absorbing resin is sent to the outlet of the sorting and feeding unit; the outlet of the sorting and feeding unit is connected to a linear feeding track. An opening for inserting a metal pipe is provided on the side wall of the detachable printing nozzle. The end of the linear feeding track is fixed near the opening of the conical nozzle of the detachable printing nozzle through a connecting piece. One end of the metal pipe is connected to the end of the linear feeding track, and the other end extends into the detachable printing nozzle and is coaxial with the nozzle of the detachable printing nozzle. There is a gap between the metal pipe and the nozzle of the detachable printing nozzle. A CCD camera is arranged at the outlet of the sorting and feeding unit for photographing and identifying the materials at the outlet of the sorting and feeding unit. A waste material dropping port is arranged on the linear feeding track near the outlet of the sorting and feeding unit. The opening and closing of the waste material dropping port is controlled by an electric valve. When it is identified that the material at the outlet of the sorting and feeding unit is not a complete spherical shape, the electric valve is controlled to open, so that the material falls into the waste material dropping port; if it is identified that the material is a complete spherical shape, the electric valve is not opened, and the material rolls into the metal pipe along the linear feeding track. A waste material bucket is arranged below the waste material dropping port. The water weigher, the dry resin storage unit, the water absorption unit, the electric valve, the sorting and feeding unit, and the CCD camera are all electrically connected to the control unit. The target recognition algorithm is loaded in the control unit and can quickly recognize the shape of the wet water-absorbing resin. If it is recognized as a complete spherical wet water-absorbing resin, it is directly sent to the metal pipe through the linear feeding track; if it is recognized as an irregular or broken wet water-absorbing resin, the electric valve at the position of the waste material dropping port is opened to make the material fall and enter the waste material bucket.

2. The device according to claim 1, characterized in that, The target recognition algorithm can quickly recognize the shape of the material within 0.3s - 0.5s. The materials reach the outlet position of the sorting and feeding unit one by one for individual recognition. After the previous material leaves the position of the waste material dropping port, the control unit closes the electric valve and then judges whether to open the electric valve again according to the recognition result of the next material. The interval between adjacent two materials is adjusted according to the adjacent resin printing gap required in actual 3D printing, ensuring that each material passes through the judgment in turn and only the complete spherical materials enter the metal pipe.

3. The device according to claim 1, characterized in that, The target recognition algorithm is an improved YOLOv7 deep neural network. The improved YOLOv7 deep neural network includes a backbone network Backbone and a Head network. The input image is input into the backbone network Backbone, and in the backbone network Backbone, it passes through CBS1, CBS2, CBS1, CBS2, ELAN, MP-1, ELAN, MP-1, ELAN, MP-1, ELAN in sequence and then outputs SPPCSPC connecting to the Head network. SPPCSPC passes through a CBS3 and Upsample to output result A; the output of the penultimate ELAN is simultaneously spliced with result A through a CBS3, and the spliced result passes through the first ELAN-W, CBS3, and Upsample to output result B; the output of the second ELAN is spliced with result B through a CBS3, and the spliced result passes through the second ELAN-W and is simultaneously connected to MP-2 and REP. The output of MP-2 is spliced with the output of the first ELAN-W processing of result A. After the spliced output is processed by the third ELAN-W, it is spliced with the output of SPPCSPC through another MP-2, and finally processed by the fourth ELAN-W. The outputs of the second, third, and fourth ELAN-W are all connected to a CBM block through a REP, and the outputs of the three CBM blocks are connected to each other to output a large-pixel feature map as the output image of the final network.

4. The device according to claim 3, characterized in that, The structures of CBS1, CBS2, and CBS3 are the same, and each includes a convolutional layer, an LN layer, and a DSiLU function. CBS1: k = 3, s = 1; CBS2: k = 3, s = 2; CBS3: k = 1, s = 1, where k is the convolutional kernel size and s is the stride. REP is composed of a convolutional layer and an LN layer in series to form one branch. There are two such identical branches in parallel and then in parallel with a branch composed of an LN layer. The outputs of the three branches are processed by addition to obtain the output of REP. MP-1 and MP-2 have the same structure, and each includes two branches. One branch is composed of a max-pooling layer and a CBS3 connection, and the other branch is composed of a CBS3 and CBS2 connection. The outputs of the two branches are spliced to obtain the output of MP-1 or MP-2. Among them, the number of hidden layers of MP-1 is i / 2, and the number of hidden layers of MP-2 is i, where i represents the number of input nodes. The structure of SPPCSPC is: the data passes through the CBS3, CBS1, CBS3 modules once and then undergoes three max-pooling processes. The results of the three max-pooling processes are spliced with the output of the last CBS3. After that, it passes through a CBS3 and is directly spliced with the output of the input of SPPCSPC processed by a CBS3. Finally, the output of SPPCSPC is obtained after passing through a CBS3. The input of ELAN passes through a max pooling layer and the output result E of CBS3. The input of ELAN also sequentially passes through CBS3, CBS1, CBS1, CBS1 to obtain the output result F. Then, the results G, H, I obtained by only passing through CBS3, CBS3 and CBS1, CBS3 and two CBS1 respectively are used. Furthermore, E, F, G, H, I are jointly subjected to a splicing operation, and finally the output of ELAN is obtained through CBS3; ELAN-W is similar in structure to ELAN. The difference is that there are four CBS1 on the result F branch. The input of ELAN sequentially passes through CBS3, CBS1, CBS1, CBS1, CBS1 to obtain the output result F. Then, the results G, H, I, J obtained by only passing through CBS3, CBS3 and CBS1, CBS3 and two CBS1, CBS3 and three CBS1 respectively are used. Furthermore, E, F, G, H, I, J are jointly subjected to a splicing operation, and finally the output of ELAN is obtained through CBS3; The CBM block is composed of a convolutional layer, an LN layer and a sigmod function.

5. The device according to claim 1, characterized in that, The end of the metal pipe is lower than the nozzle end of the detachable printing nozzle by a distance equal to the diameter of a spherical water-absorbing resin, so that the concrete slurry wraps the spherical water-absorbing resin before exiting the nozzle of the detachable printing nozzle and prints together; or the end of the metal pipe is located 1-2 mm above the nozzle end of the detachable printing nozzle.

6. The device according to claim 1, characterized in that, An electromagnetic valve (11) is provided at the connection position between the linear feeding track and the metal pipe, which is used to control whether to lay the resin material for the concrete. If the laying is not carried out, the equipment for resin conveying is controlled to stop working.

7. The device according to claim 1, characterized in that, The particle size range of the dry water-absorbing resin is 1-2 mm. The dry water-absorbing resin is first saturated with water and then extruded together with the concrete at the nozzle position. The water-absorbing resin is fed into the detachable printing nozzle of the spiral extrusion unit and extruded together with the concrete. The outlet of the metal pipe faces the axis of the detachable printing nozzle.

8. The device according to claim 1, characterized in that, A grating counter is provided below the position of the waste material falling inlet, and the grating counter is connected to the control unit.

Images