Underwater sucking and catching device based on spring force storage device and dynamic modeling method

Abstract

The invention relates to an underwater suction catching device based on a spring force storage device and a dynamic modeling method, and solves the technical problem of how to reduce damage to sea cucumbers in the sea cucumber catching process. The device comprises a collection box, a water suction pipeline, a drainage pipeline, a communication base, a front supporting plate, a rear supporting plate, a left fixing rod, a right fixing rod, a lead screw motor, a sealing cover, a cylinder, a pushing shaft, a ferromagnetic metal disc, a spring and an annular electromagnet. The sea cucumber catching task is completed with high efficiency, high quality and high safety in a soft catching mode. The invention belongs to the technical field of cross fusion of ocean engineering equipment and aquatic product fishing machinery.

Description

Technical Field

[0001] This invention relates to the field of sea cucumber harvesting technology, and more specifically, to an underwater suction device based on a spring-powered storage device and a dynamic modeling method. Background Technology

[0002] Oceans cover 70.8% of the Earth's surface and contain abundant mineral and marine biological resources, which are important assets for the sustainable development of human society.

[0003] Sea cucumbers are a highly nutritious seafood product. They feed on seaweed and plankton and are typically farmed in nearshore waters. Currently, sea cucumber harvesting mainly involves manual operations using simple tools through diving or trawling, which are inefficient, costly, and dangerous. Existing technologies also include harvesting methods using robotic arms to grasp sea cucumbers (see patent applications CN118489638A and CN112544577A), but these robotic arms can easily damage sea cucumbers and result in low operational efficiency.

[0004] Therefore, how to reduce damage to sea cucumbers is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] This invention aims to solve the technical problem of how to reduce damage to sea cucumbers during the harvesting process. It provides an underwater suction device based on a spring-powered storage device and a dynamic modeling method that reduces damage to sea cucumbers, achieving high efficiency, high quality, and high safety.

[0006] This invention provides an underwater suction device based on a spring-powered storage mechanism, comprising a collection box, a suction pipe, a drainage pipe, a connecting seat, a front support plate, a rear support plate, a left fixed rod, a right fixed rod, a screw motor, a sealing cover, a cylinder, a push shaft, a ferromagnetic metal disc, a spring, and a ring electromagnet. The collection box includes a cover plate, which is detachably connected to the bottom of the collection box. The connecting seat is fixedly connected to the top of the collection box, and the lower part of the suction pipe is fixedly connected to the top of the connecting seat. The connecting seat has a channel, and the lower part of the suction pipe has an opening that communicates with the channel of the connecting seat. The top of the collection box has an inlet that communicates with the channel of the connecting seat. The drainage pipe is connected to the rear of the suction pipe. The front support plate is fixedly connected to the rear end of the suction pipe. The left fixed rod is fixedly connected between the front support plate and the rear support plate, and the right fixed rod is fixedly connected between the front support plate and the rear support plate. The screw motor is fixedly connected to the rear support plate. The sealing cover is... The end of the lead screw motor is connected to a sealing cover, and a sealing ring is connected between the end of the lead screw motor and the sealing cover. The rear end of the cylinder is fixedly connected to the sealing cover, and the front end of the cylinder is fixedly connected to the front support plate. The front support plate has a central through hole, which communicates with the cylinder. The lead screw motor has a lead screw, and the sealing cover has a central through hole. The lead screw passes through the central through hole of the sealing cover and is located in the cylinder. The push shaft has an axial internal thread hole, which is connected and engaged with the lead screw. The push shaft passes through the central through hole of the front support plate and can move back and forth in the water suction pipe. A ferromagnetic metal disc is located in the water suction pipe, with its edge close to the inner wall of the water suction pipe. A spring is located in the water suction pipe, with its front end fixedly connected to the ferromagnetic metal disc and its rear end fixedly connected to the front support plate. A ring electromagnet is fixedly connected to the outside of the water suction pipe. The ring electromagnet includes a ring iron core and a coil, with the coil wound on the ring iron core.

[0007] Preferably, when the lead screw of the lead screw motor rotates, the push shaft moves forward, and the front end of the push shaft pushes the ferromagnetic metal disc forward to the position of the annular electromagnet. The ferromagnetic metal disc is located in front of the opening at the bottom of the water suction pipe. The spring is stretched, and the annular electromagnet attracts the ferromagnetic metal disc when energized. The lead screw rotates, causing the push shaft to move backward. The front end of the push shaft moves to the position of the front support plate. At this time, the marine organism capture system based on the spring storage device is in the initial state. In the initial state, the annular electromagnet is de-energized, and under the tension of the spring, the ferromagnetic metal disc moves backward to the position of the drain pipe.

[0008] Preferably, the drainage pipe includes a left drainage pipe and a right drainage pipe, with the left drainage pipe connected to the rear of the water intake pipe and the right drainage pipe connected to the rear of the water intake pipe.

[0009] Preferably, the left and right drainage pipes are arranged symmetrically.

[0010] Preferably, the spring is a spring that surrounds the push shaft.

[0011] Preferably, there are two or more springs, which are located around the push shaft, and the push shaft can move back and forth in the space between the two or more springs.

[0012] Preferably, the spring includes a first spring, a fourth spring, a second spring, and a third spring.

[0013] The present invention also provides a sea cucumber suction robot, including a main frame, on which any of the aforementioned underwater suction devices based on a spring-powered device are connected.

[0014] This invention also provides a dynamic modeling method for an underwater suction device based on a spring-powered accumulator, comprising the following steps:

[0015] A ferromagnetic metal disk is defined as the piston, a water suction pipe as the pipe, and the front support plate, rear support plate, left fixed rod, right fixed rod, screw motor housing, sealing cover, and cylinder as the support. A complete dynamic model including spring, piston, pipe, and support is constructed, and the physical information neural network PINN model is introduced to perform high-precision identification and joint solution of key fluid dynamic parameters.

[0016] When the toroidal electromagnet is de-energized, the four pre-stretched springs simultaneously pull the piston, causing it to accelerate in the seawater. The axial dynamic equation is:

[0017]

[0018] In the formula, M is the piston mass, 4k is the total stiffness coefficient of the four springs, x0 is the initial tension of the springs, and x is the axial displacement of the piston relative to its initial position. The axial velocity of the piston; Where C is the piston cross-sectional area, D is the diameter of the suction pipe, and L is the length of the suction pipe, where L / D > 5; d ρ is the equivalent drag coefficient of the piston; ρ is the density of seawater. For fluid viscous resistance; reaction force F accel (t) and F pressure (t) is obtained from the output of the physical information neural network PINN, F pressure The value of (t) is equal to the suction force F at the suction working surface. suction (t);

[0019] reaction force F accel (t) is:

[0020]

[0021] In the formula, L effThe equivalent acceleration length is given by α, where α is the velocity distribution correction factor, and v f (t) represents the average flow velocity inside the pipe;

[0022] Suction F suction (t) is:

[0023]

[0024] The process of constructing the PINN (Physical Information Neural Network) model is as follows:

[0025] Step (1), definition of physical governing equations and solution domain:

[0026] Navier–Stokes equations:

[0027]

[0028] In the formula, u=(u r ,u θ ,u x ) represents the velocity vector, and p represents the pressure.

[0029] Solution domain: includes the inlet region in front of the piston, the inside of the pipe, and the surrounding flow field region outside the pipe. The piston surface constitutes a time-varying motion boundary, while the inlet, outlet, and pipe wall are fixed boundaries.

[0030] Step (2), Network Structure and Input / Output Design:

[0031] Input quantities: spatial coordinates (x, r, θ), time t, and the axial displacement x(t) of the piston relative to its initial position and the axial velocity v of the piston. piston (t);

[0032] Output quantities: Local pressure and velocity fields [p(x,r,θ,t),u] r ,u θ ,u x ], and a set of equivalent parameter vectors η to be identified:

[0033] η=[α,C L ,L eff ,ζ]

[0034] Step (3), the total loss function is defined as:

[0035] L=λ1L PDE +λ2L BC +λ3L data

[0036] Among them, the physical residual loss is in the internal sampling point set Ω int Above, definition:

[0037]

[0038] In the formula, The differential operator representing the Navier–Stokes equations;

[0039] λ1, λ2, λ3 are weighting coefficients;

[0040] Step (4), training and parameter identification process;

[0041] Step S401: Obtain spatial coordinates (x, r, θ), time t, and axial displacement x of the piston relative to its initial position. (0) (t), piston axial velocity

[0042] Step S402: In the corresponding spatiotemporal domain, internal points and boundary points are generated using Latin hypercube or random sampling to form a training set;

[0043] Step S403, using spatial coordinates (x, r, θ), time t, and the axial displacement x of the piston relative to its initial position. (0) (t) and piston axial velocity As input, construct PINN and minimize the loss function L in the above equation to obtain the local pressure p, velocity field u, and equivalent parameter vector η;

[0044] η=[α,C L ,L eff ,ζ]

[0045] Step S404: Calculate each fluid force term using the identified parameters:

[0046] Calculate the reaction force F accel (t):

[0047]

[0048] In the formula, the average flow velocity v in the pipe f (t) is: v f (t)≈v piston (t);

[0049] Calculate the suction force F suction (t):

[0050]

[0051] Calculate the equivalent lateral force F lateral (t):

[0052]

[0053] Then, the piston motion is resolved using the following axial dynamic equations to obtain x.(1) (t),

[0054]

[0055] Next, the relative error of the displacement trajectory is calculated. If E exceeds the preset convergence threshold ∈ = 10 -3 Then determine x (1) (t) and the initial input x (0) If there is a significant difference in (t), then x will be... (1) (t), As new input, repeat steps S402-S404 for iterative training until E≤∈, that is, the flow field output by PINN converges with the motion state predicted by the dynamic equation.

[0056] Preferably, the structure of the suction device is optimized:

[0057] Step (S1): Manually define the spring stiffness k, pre-tension x0, piston mass M, suction pipe length L, and bracket installation length L. pipe The specific values, combined with spatial coordinates (x, r, θ), time t, axial displacement x(t) of the piston relative to its initial position, and axial velocity v of the piston, are determined by... piston (t), forming a set of data, then the spatial coordinates (x,r,θ), time t, axial displacement x(t) of the piston relative to the initial position, and axial velocity v of the piston are... piston (t) is input to the PINN model, and the PINN model outputs the local pressure p(x,r,θ,t) and the velocity field [u r ,u θ ,u x ],α,C L ,L eff ,ζ;

[0058] Step (S2): Calculate the reaction force F using the following formula. accel (t):

[0059]

[0060] Calculate the suction force F using the following formula. suction (t):

[0061]

[0062] The effective lateral force F is calculated using the following formula. lateral (t):

[0063]

[0064] Step (S3): Calculate key performance indicators:

[0065] The maximum equivalent stress σ of the stent is calculated using the following formula. max :

[0066]

[0067] In the formula, M bend (t)=F lateral (t)·L pipe c = D / 2

[0068] The maximum suction force F is calculated using the following formula. suction,max :

[0069] F suction,max =max(F suction (t))

[0070] With the maximum equivalent stress σ of the stent max Under the constraint of not exceeding the allowable stress [σ] of the support material, the peak instantaneous suction force F is made to capture. suction,max To maximize the target, the gradient descent method is used to find the optimal value using the automatic differentiation property of the PINN framework. The steps are as follows:

[0071] Step 1), define the optimization objective function and construct the loss function J that includes the objective term and the penalty term. opt :

[0072] J opt =-F suction,max +w·ReLU(σ max -[σ])

[0073] In the formula, the first term is the negative value of the maximum suction force; the second term is the penalty term for stress constraint, w is the penalty weight, and ReLU is the linear rectification function;

[0074] Step 2), calculate the objective function J opt For design variables (k, x0, L, D, L) pipe gradient of )

[0075] Step 3), variable update: iteratively update the design variables along the negative gradient direction.

[0076]

[0077] In the formula, X=[k,x0,L,D,L pipe ] T The design variable vector is given, where γ is the learning rate;

[0078] Step 4) Iteratively solve the problem by repeating the gradient calculation and variable update steps above until the design variables converge, thereby obtaining the optimal combination of structural parameters that satisfies the strength constraints and maximizes the suction force.

[0079] The advantages of this invention are its innovative suction method, improved suction effect, and flexible, convenient, and efficient operation. It uses negative pressure suction to capture sea cucumbers, minimizing damage and ensuring high operational efficiency. A controllable negative pressure is generated through a spring-loaded system, allowing for precise and efficient suction of the sea cucumbers.

[0080] Flexible in use, suitable for various seabed terrains and complex sea conditions. Enables efficient, high-quality, and highly safe sea cucumber harvesting.

[0081] It minimizes disturbance to the seabed sedimentary ecosystem and enables the immediate separation of seawater, sediment, and debris.

[0082] Its structure is relatively simple, it is easier to operate, and its manufacturing and maintenance costs are relatively low.

[0083] This invention is not limited to absorbing sea cucumbers, but can also absorb other marine organisms.

[0084] Further features and aspects of the present invention will be clearly described in the following detailed description with reference to the accompanying drawings. Attached Figure Description

[0085] Figure 1 This is an isometric view of the suction device;

[0086] Figure 2 yes Figure 1 Front view of the suction device shown;

[0087] Figure 3 yes Figure 1 Rear view of the suction device shown;

[0088] Figure 4 yes Figure 1 Right view of the suction device shown;

[0089] Figure 5 yes Figure 1 Left view of the suction device shown;

[0090] Figure 6 yes Figure 1 Top view of the suction device shown;

[0091] Figure 7 yes Figure 6 A cross-sectional view along the AA direction in the structure shown;

[0092] Figure 8 yes Figure 1Another isometric view of the suction device shown;

[0093] Figure 9 yes Figure 1 Another isometric view of the suction device shown;

[0094] Figure 10 yes Figure 1 The diagram shows the installation of the four springs in the suction device.

[0095] Figure 11 yes Figure 1 The diagram shows the installation of the four springs in the suction device.

[0096] Figure 12 This is a diagram showing the state of the suction device with its four springs in a compressed state and the ferromagnetic metal disc located at the left-side drain pipe position.

[0097] Figure 13 yes Figure 12 Top view of the structure shown;

[0098] Figure 14 yes Figure 13 Cross-sectional view along the BB direction;

[0099] Figure 15 yes Figure 12 The diagram shows the state of the ferromagnetic metal disk and the four springs in the structure shown.

[0100] Figure 16 yes Figure 15 Front view of the structure shown;

[0101] Figure 17 This is a diagram showing the dimensions of each part of the suction device.

[0102] Explanation of symbols in the diagram:

[0103] 1. Collection box, 1-1. Cover plate, 2. Water suction pipe, 3. Left drainage pipe, 4. Right drainage pipe, 5. Connecting seat, 6. Front support plate, 7. Rear support plate, 8. Left fixing rod, 9. Right fixing rod, 10. Screw motor, 10-1. Screw, 11. Sealing cover, 12. Sealing ring, 13. Cylinder, 14. Push shaft, 15. Ferromagnetic metal disc, 16. First spring, 17. Fourth spring, 18. Second spring, 19. Third spring, 20. Ring electromagnet. Detailed Implementation

[0104] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0105] The specific embodiments described below are merely preferred embodiments of this application, and the scope of protection of this application is not limited thereto. Those skilled in the art can make modifications or variations based on the principles, concepts, and spirit of this application, and the resulting technical solutions should all be covered within the scope of protection of this application.

[0106] Example 1

[0107] like Figure 1-11As shown, the suction device includes a collection box 1, a water suction pipe 2, a left drainage pipe 3, a right drainage pipe 4, a connecting seat 5, a front support plate 6, a rear support plate 7, a left fixing rod 8, a right fixing rod 9, a lead screw motor 10, a sealing cover 11, a sealing ring 12, a cylinder 13, a push shaft 14, a ferromagnetic metal disc 15, a first spring 16, a fourth spring 17, a second spring 18, a third spring 19, and a ring electromagnet 20. The collection box 1 includes a cover plate 1-1, which is detachably connected to the bottom of the collection box 1. The connecting seat 5 is fixedly connected to the top of the collection box 1. The lower part of the water suction pipe 2 is fixedly connected to the top of the connecting seat 5. The connecting seat 5 has a channel, and the lower part of the water suction pipe 2 has an opening that communicates with the channel of the connecting seat 5. The top of the collection box 1 has an inlet that communicates with the channel of the connecting seat 5. The left drainage pipe 3 is connected to the rear of the water suction pipe 2, and the right drainage pipe 4 is connected to the rear of the water suction pipe 2. The front support plate 6 is fixedly connected to the rear end of the water suction pipe 2. The front end of the left fixing rod 8 is fixedly connected to the front support plate 6, and the rear end of the left fixing rod 8 is fixedly connected to the rear support plate 7. The front end of the right fixing rod 9 is fixedly connected to the front support plate 6, and the rear end of the right fixing rod 9 is fixedly connected to the rear support plate 7. The lead screw motor 10 is fixedly installed on the rear support plate 7. The sealing cover 11 is connected to the end of the lead screw motor 10. A sealing ring 12 is connected between the sealing cover 11 and the end of the lead screw motor 10. The rear end of the cylinder 13 is fixedly connected to the sealing cover 11, and the front end of the cylinder 13 is fixedly connected to the front support plate 6. The front support plate 6 has a central through hole, which communicates with the cylinder 13. The output part of the lead screw motor 10 is the lead screw 10-1. The sealing cover 11 has a central through hole, through which the lead screw 10-1 passes. The lead screw 10-1 is located in the cylinder 13. The push shaft 14 has an axial internal thread hole, which connects and engages with the lead screw 10-1 to achieve the connection between the push shaft 14 and the lead screw 10-1. The push shaft 14 passes through the central through hole of the front support plate 6. A ferromagnetic metal disc 15 is located in the water suction pipe 2, with its edge close to the inner wall of the water suction pipe 2. A first spring 16, a second spring 18, a third spring 19, and a fourth spring 17 are all located in the water suction pipe 2. The front end of the first spring 16 is fixedly connected to the ferromagnetic metal disc 15, and the rear end of the first spring 16 is fixedly connected to the front support plate 6. The front end of the fourth spring 17 is fixedly connected to the ferromagnetic metal disc 15, and the rear end of the fourth spring 17 is fixedly connected to the front support plate 6. The front end of the second spring 18 is fixedly connected to the ferromagnetic metal disc 15, and the rear end of the second spring 18 is fixedly connected to the front support plate 6. The front end of the third spring 19 is fixedly connected to the ferromagnetic metal disc 15, and the rear end of the third spring 19 is fixedly connected to the front support plate 6.The first spring 16, the second spring 18, the third spring 19, and the fourth spring 17 are located around the push shaft 14, meaning the push shaft 14 can move back and forth within the space between the first spring 16, the second spring 18, the third spring 19, and the fourth spring 17. The annular electromagnet 20 is fixedly installed on the outside of the water suction pipe 2. The annular electromagnet 20 includes an annular iron core and a coil, with the coil wound around the annular iron core. Figure 1-11 The display shows that the rotation of the lead screw 10-1 causes the push shaft 14 to move forward. The front end of the push shaft 14 pushes the ferromagnetic metal disk 15 forward to the position of the annular electromagnet 20. The ferromagnetic metal disk 15 is located in front of the opening at the bottom of the water suction pipe 2. The first spring 16, the second spring 18, the third spring 19 and the fourth spring 17 are all stretched. The annular electromagnet 20 attracts the ferromagnetic metal disk 15 when it is energized.

[0108] exist Figure 1-11 Based on the state of the suction device shown, the lead screw motor 10 operates to rotate the lead screw 10-1, thereby moving the push shaft 14 backward. The front end of the push shaft 14 moves to the position of the front support plate 6 (e.g., Figure 14 As shown), at this time, the pushing force of the front end of the push shaft 14 on the ferromagnetic metal disk 15 disappears. Since the annular electromagnet 20 attracts the ferromagnetic metal disk 15, the ferromagnetic metal disk 15 is still positioned at the position of the annular electromagnet 20. Next, the power to the annular electromagnet 20 is turned off, the magnetic force of the annular electromagnet 20 disappears, and under the pulling force of the four springs, the ferromagnetic metal disk 15 quickly moves backward to the position of the left drainage pipe 3 and the right drainage pipe 4, as shown. Figure 12 and 14 As shown.

[0109] When using the aforementioned suction device to harvest sea cucumbers, the device is placed on the seabed or mounted on the main frame of an underwater robot. The suction process is as follows:

[0110] The first step is to put the suction device in its initial state. The lead screw motor 10 operates, causing the lead screw 10-1 to rotate, thereby moving the push shaft 14 forward. The front end of the push shaft 14 pushes the ferromagnetic metal disc 15 forward to the position of the annular electromagnet 20. The ferromagnetic metal disc 15 is located in front of the opening at the bottom of the water suction pipe 2. The first spring 16, the second spring 18, the third spring 19, and the fourth spring 17 are all stretched, energizing the annular electromagnet 20. The annular electromagnet 20 attracts and positions the ferromagnetic metal disc 15. Then, the lead screw 10-1 rotates in the opposite direction, causing the push shaft 14 to move backward, moving the front end of the push shaft 14 to the position of the front support plate 6. The ferromagnetic metal disc 15 is still positioned at the position of the annular electromagnet 20. At this time, the suction device is in its initial state.

[0111] The second step is to align the front end of the water intake pipe 2 with the sea cucumber area.

[0112] The third step is to de-energize the annular electromagnet 20. Under the tension of the four springs, the ferromagnetic metal disc 15 quickly moves backward to the position of the left drainage pipe 3 and the right drainage pipe 4. Seawater, along with sea cucumbers, is sucked in from the front port of the water intake pipe 2. The seawater flows backward along the water intake pipe 2 and is discharged outward from the left drainage pipe 3 and the right drainage pipe 4 (the ferromagnetic metal disc 15 is similar to a piston). The sea cucumbers fall down into the collection box 1 from the opening at the bottom of the water intake pipe 2 through the channel of the connecting seat 5 and then from the entrance at the top of the collection box 1, completing the sea cucumber collection.

[0113] Repeat steps one, two, and three above to conduct the next data collection.

[0114] After multiple collection operations, the suction device is transferred from the seabed to the shore or a boat, the cover plate 1-1 is removed, and then the sea cucumbers in the collection box 1 can be taken out.

[0115] As can be seen, the controllable negative pressure generated by the spring storage system can accurately and efficiently extract sea cucumbers, reducing damage to them, minimizing disturbance to the seabed sedimentary ecosystem, and achieving immediate separation of sea cucumbers from mud and debris. The mud and debris are discharged from the left drainage pipe 3 and the right drainage pipe 4.

[0116] It should be noted that the symmetrical arrangement of the left drain pipe 3 and the right drain pipe 4 on both sides of the rear of the water intake pipe 2, as shown in the attached diagram, is the optimal design. Alternatively, a single drain pipe can also be installed.

[0117] It should be noted that instead of four springs, only one spring can be used, which surrounds the push shaft 14, meaning the push shaft 14 is located in the middle space of this single spring. Of course, using only one spring, such as the first spring 16, is also feasible; or using two springs (two of the first spring 16, the fourth spring 17, the second spring 18, and the third spring 19) is also feasible; or even using three springs (three of the first spring 16, the fourth spring 17, the second spring 18, and the third spring 19) is also feasible.

[0118] The aforementioned suction device, as a marine organism suction system based on a spring-powered device, is not limited to suction of other organisms.

[0119] Example 2

[0120] The structural design of the suction device in Example 1 was optimized using dynamic modeling methods.

[0121] During underwater operation, the rapid piston movement of the suction device induces fluid flow, generating complex forces on the suction pipe 2 and the mounting body (such as an underwater robot). Therefore, a mathematical model is established to analyze this coupled dynamic process, and a Physics-Informed Neural Network (PINN) is directly embedded into the dynamic modeling and solution.

[0122] A complete dynamic model was constructed, including a spring, piston (ferromagnetic metal disk), pipe (water suction pipe 2), and support. A Physical Information Neural Network (PINN) was introduced to accurately identify and jointly solve key fluid dynamic parameters. By integrating PINN into the dynamic modeling and parameter inversion process, the dynamic model maintains the interpretability of the analytical structure while possessing high-precision prediction capabilities for complex nonlinear flows and boundary conditions, thus providing a quantitative basis for the structural optimization of the device.

[0123] When the annular electromagnet 20 is de-energized, the four pre-stretched springs simultaneously release their stored elastic energy, pulling the piston (ferromagnetic metal disk) to accelerate in the seawater. During its motion, the piston is subjected to the combined effects of various forces, including the spring restoring force, fluid viscous resistance, pressure difference force, and fluid acceleration reaction force. Its axial dynamic equation can be uniformly written as:

[0124]

[0125] In the formula, M is the piston mass. 4k is the total stiffness coefficient of the four springs. x0 is the initial tension of the spring (reference). Figure 17 x is the axial displacement of the piston relative to its initial position. This represents the axial velocity of the piston. Let D be the cross-sectional area of ​​the piston, and D be the diameter of the suction pipe 2. The length of the suction pipe 2 is L, and L / D > 5. d The equivalent drag coefficient of the piston can be determined through a water tank experiment or obtained by looking up the Reynolds number in a standard drag coefficient database. F represents the viscous drag of the fluid. ρ is the density of seawater. accel (t) and F pressure (t) is given by the subsequent PINN fluid field solution, achieving unification with the fluid induced force model. F pressure The value of (t) is equal to the suction force F at the suction working surface. suction (t).

[0126] In the initial stage of piston acceleration, due to the small fluid resistance, the influence of fluid forces can be ignored. At this time, the system degenerates into simple harmonic motion, with the characteristic frequency being:

[0127]

[0128] Given k, M, and initial conditions, the peak piston velocity v can be obtained. max The time and its corresponding time are used as boundary conditions and reference time scales when modeling PINN.

[0129] As the piston moves forward, the seawater inside the pipe must be accelerated to maintain continuity, generating an accelerating reaction force on the piston. For consistency with PINN representation, this reaction force is written in an equivalent form:

[0130]

[0131] In the formula, ρ is the density of seawater, which can be considered a constant under given water depth and temperature conditions. L eff The equivalent acceleration length is given by α, where α is the velocity distribution correction factor, and v f (t) represents the average flow velocity within the pipe. These coefficients were identified by PINN using the Navier–Stokes equations in conjunction with a small amount of data to avoid systematic errors caused by one-dimensional simplification.

[0132] F after PINN calibration can be used accel The (t) curve is used to perform sensitivity analysis on the piston acceleration process under different spring parameters and pipe geometry combinations, so as to reasonably select L. eff The pipe diameter D and spring stiffness ensure that the acceleration reaction force is neither too large, affecting the response speed, nor too large, causing structural overload. Simultaneously, the piston cross-sectional area A = πD is directly obtained from D. 2 / 4.

[0133] When the piston moves at high speed, a relatively low-pressure area is formed in front of it, causing the front end of the pipe (the working surface of the suction device) to be subjected to a forward suction force, approximately as follows:

[0134]

[0135] v given by PINN piston (t) and the local pressure field distribution can be used to inversely deduce the target suction force F. suction,target The required spring pretension x0 and spring stiffness k are used to quantify the "capture capacity" into a designable combination of structural parameters.

[0136] The high-speed piston motion generates a flow field around it, which induces a lateral force on the mounting bracket. To correspond with the PINN output, the equivalent lateral force in the suction pipe 2 is:

[0137]

[0138] In the formula, C L The equivalent lift coefficient, The geometric factor is used. The three-dimensional flow field is constrained and calibrated using PINN, and C is obtained through inversion. L It can be directly used for structural stress analysis.

[0139] F accurately predicted using PINN lateral,max Based on the time of its action, the cross-sectional dimensions of the support, material strength, and installation length L can be directly determined. pipe Through verification and optimization, a slender support design was achieved that is "neither overly conservative nor overloaded to the point of failure".

[0140] Therefore, the process of constructing the Physical Information Neural Network (PINN) model is as follows:

[0141] Step (1), definition of physical governing equations and solution domain:

[0142] Navier–Stokes equations:

[0143]

[0144] In the formula, u=(u r ,u θ ,u x ) represents the velocity vector, and p represents the pressure.

[0145] Solution domain: includes the inlet region in front of the piston, the inside of the pipe, and the surrounding flow field region outside the pipe. The piston surface constitutes a time-varying motion boundary, while the inlet, outlet, and pipe wall are fixed boundaries.

[0146] Step (2), Network Structure and Input / Output Design:

[0147] Input quantities: spatial coordinates (x, r, θ), time t, and piston motion state: axial displacement x(t) of the piston relative to its initial position, and axial velocity v of the piston. piston (t). x(t),v during the training phase piston (t) Data can be obtained in two ways: the first is the numerical simulation results based on the dynamic equations (which give the prior motion trajectory and are used to construct a physically consistent initial flow field), and the second is the actual measurement data of the piston stroke through displacement sensors or encoders (the velocity is obtained by numerically differentiating the displacement signal).

[0148] Output quantities: Local pressure and velocity fields [p(x,r,θ,t),u] r ,u θ ,u x ], and a set of equivalent parameter vectors to be identified.

[0149] η=[α,C L ,L eff ,ζ,…]

[0150] Here, η represents a set of equivalent parameters independent of spatial location and time (such as inlet correction coefficient, lift coefficient, equivalent acceleration length, and equivalent damping ratio), which should be considered as fixed constants under a given device structure and operating conditions. In PINN, these constants are automatically determined as globally trainable parameters along with the network weights by minimizing the total loss function, thus remaining consistent throughout the solution domain and time interval.

[0151] On the other hand, under most operating conditions, it is difficult to deploy a large number of pressure and velocity sensors throughout the entire flow field to directly obtain p and u at every spatial point. Instead, the data loss term L is constructed using only a small number of readily available observables (such as the pressure difference between the inlet and outlet, the total flow rate in the pipe, strain or stress on the support, etc.). data Under the physical constraints of the Navier–Stokes equations and boundary conditions, PINN infers the pressure distribution, velocity distribution, and optimal equivalent parameter vector η of the entire flow field by simultaneously minimizing the equation residuals, boundary condition residuals, and data residuals, without requiring point-by-point measurements at every spatial location.

[0152] Step (3), the total loss function is defined as:

[0153] L=λ1L PDE +λ2L BC +λ3L data

[0154] Among them, the physical residual loss is in the internal sampling point set Ω int Above, definition:

[0155]

[0156] In the formula, This represents the differential operator of the Navier–Stokes equations.

[0157] Boundary condition loss, defined on the set Γ of boundary sampling points such as pipe wall, inlet / outlet, and piston surface, under constraints of no slip or given pressure and flow conditions:

[0158]

[0159] Data consistency loss, defined as the constraint imposed on observables (such as inlet pressure difference Δp(t), total flow rate Q(t), and support stress σ(t)) on a small set of experimental or high-precision numerical data points D, is as follows:

[0160]

[0161] The weighting coefficients λ1, λ2, and λ3 are used to assess the importance of the residuals of the equilibrium equations, boundary conditions, and observed data.

[0162] Step (4), training and parameter identification process.

[0163] Step S401: Obtain spatial coordinates (x, r, θ), time t, and axial displacement x of the piston relative to its initial position. (0) (t), piston axial velocity

[0164] Step S402: In the corresponding spatiotemporal domain, internal points and boundary points are generated using Latin hypercube or random sampling to form a training set.

[0165] Step S403, using spatial coordinates (x, r, θ), time t, and the axial displacement x of the piston relative to its initial position. (0) (t) and piston axial velocity As input, PINN is constructed and the loss function L in the above equation is minimized to obtain the local pressure p, velocity field u, and equivalent parameter vector η.

[0166] η=[α,C L ,L eff ,ζ]

[0167] Step S404: Calculate each fluid force term using the identified parameters:

[0168] Calculate the reaction force F accel (t):

[0169]

[0170] In the formula, the average flow velocity v in the pipe f (t) is: v f (t)≈v piston (t).

[0171] Calculate the suction force F suction (t):

[0172]

[0173] Calculate the equivalent lateral force F lateral (t):

[0174]

[0175] Then, the piston motion is resolved using the axial dynamics equations to obtain x. (1) (t),

[0176]

[0177] Calculate the relative error of the displacement trajectory If E exceeds the preset convergence threshold ∈ = 10-3 Then determine x (1) (t) and the initial input x (0) If there is a significant difference in (t), then x will be... (1) (t), As new input, repeat steps S402-S404 for iterative training until E≤∈, that is, the flow field output by PINN converges with the motion state predicted by the dynamic equation, forming a self-consistent "flow field-dynamics-equivalent parameter" solution.

[0178] Through the above PINN mathematical modeling process, the complex three-dimensional unsteady flow and nonlinear damping characteristics are uniformly converted into a finite number of equivalent parameters η, and systematically embedded into the spring-piston-pipe-support dynamic equations, realizing high-precision prediction from device structure to force-displacement-velocity time history.

[0179] The following describes the total axial force borne by the main body (such as an underwater robot) at different stages during the operation of the suction device.

[0180] In the initial stage of the suction device's operation (when t is small), the spring restoring force is at its maximum, while the suction and other fluid forces are relatively small, and the total axial force mainly comes from the spring. In the middle stage (when the piston speed is maximum), the fluid forces (suction, resistance, and acceleration reaction force) reach their peak, possibly equal to or greater than the spring force. In the decay stage (when t is large), the piston speed decreases, various fluid forces decrease, and eventually tend to reach a static state.

[0181] Taking into account the influence of various forces, the total axial force F axial (t) is:

[0182] F axial (t)=4k(x0-x(t))-F suction (t)-F accel (t)-F drag (t)

[0183] In the formula, This is the fluid viscous resistance. Physically, it reflects the superposition of the spring reaction force (backward) and the fluid traction force (forward) acting on the mounting body, and its magnitude is equal to the resultant force driving the piston to accelerate.

[0184] Lateral force F lateral (t) A bending moment is generated at a certain distance from the support connection point. (Reference) Figure 17 The distance from the support connection point to the suction working surface is L. pipe L pipe The actual length is the installation length of the bracket, then the bending moment that the bracket bears is:

[0185] M bend (t)=Flateral (t)·L pipe

[0186] This bending moment is the primary cause of bending strain and bending stress in the support structure. When designing the support structure, it is necessary to ensure that its bending stiffness is sufficient to withstand this bending moment M. bend (t), otherwise the stent may be subjected to excessive bending or fatigue failure.

[0187] Therefore, the trained PINN model is used to optimize the design of the suction device structure. The specific process is as follows:

[0188] Step (S1): Manually define the spring stiffness k, pre-tension x0, piston mass M, suction pipe length L, and bracket installation length L. pipe The specific values, combined with spatial coordinates (x, r, θ), time t, axial displacement x(t) of the piston relative to its initial position, and axial velocity v of the piston, are determined by... piston (t), forming a set of data, then the spatial coordinates (x,r,θ), time t, axial displacement x(t) of the piston relative to the initial position, and axial velocity v of the piston are... piston (t) is input to the PINN model, and the PINN model outputs the local pressure p(x,r,θ,t) and the velocity field [u r ,u θ ,u x ],α,C L ,L eff ,ζ.

[0189] Step (S2): Calculate the reaction force F using the following formula. accel (t):

[0190]

[0191] Calculate the suction force F using the following formula. suction (t):

[0192]

[0193] The effective lateral force F is calculated using the following formula. lateral (t):

[0194]

[0195] Step (S3): Calculate key performance indicators.

[0196] The maximum equivalent stress σ of the stent is calculated using the following formula. max :

[0197]

[0198] In the formula, M bend (t)=F lateral (t)·L pipe c = D / 2

[0199] The maximum suction force F is calculated using the following formula. suction,max :

[0200] F suction,max =max(F suction (t))

[0201] Under different simulation iterations or experimental repetitions, the suction force typically appears as several time histories, denoted as a set of time series. In this set expression, i represents the i-th sample. To provide a unified optimal metric, the global maximum attraction is defined as the global maximum value for all samples and all times:

[0202]

[0203] In the case of a single time history (N=1), the above equation is expressed as:

[0204] F suction,max =max(F suction (t)).

[0205] With "the maximum equivalent stress σ of the stent" max Under the constraint of not exceeding the allowable stress [σ] of the support material, the peak instantaneous suction force F is made to capture. suction,max The objective is to maximize the value of the product. Specifically, the gradient descent method is used to optimize the product by leveraging the automatic differentiation property of the PINN framework. The steps are as follows:

[0206] Step 1), define the optimization objective function and construct the loss function J that includes the objective term and the penalty term. opt :

[0207] J opt =-F suction,max +w·ReLU(σ max -[σ])

[0208] In the formula, the first term is the negative value of the maximum suction force (minimizing this value is equivalent to maximizing the suction force); the second term is the penalty term for stress constraints, w is the penalty weight, and ReLU is the linear rectification function (a penalty is generated when the stress exceeds the limit, otherwise it is 0).

[0209] Step 2), calculate the gradient. Since the PINN model and its dynamic equations are both composed of differentiable operations, the objective function J can be directly calculated using the backpropagation algorithm of a deep learning framework. optFor design variables (k, x0, L, D, L) pipe gradient of ) This leverages the advantage of PINN as a differentiable surrogate model, avoiding the inefficiency of traditional finite difference methods in calculating gradients.

[0210] Step 3), variable update: iteratively update the design variables along the negative gradient direction.

[0211]

[0212] In the formula, X=[k,x0,L,D,L pipe ] T The variable vector is designed, and γ is the learning rate (step size).

[0213] Step 4) Iteratively solve the problem by repeating the gradient calculation and variable update steps above until the design variables converge, thereby obtaining the optimal combination of structural parameters that satisfies the strength constraints and maximizes the suction force.

Claims

1. An underwater suction device based on a spring-powered energy storage mechanism, characterized in that, The system includes a collection box, a suction pipe, a drainage pipe, a connecting seat, a front support plate, a rear support plate, a left fixed rod, a right fixed rod, a lead screw motor, a sealing cover, a cylinder, a push shaft, a ferromagnetic metal disc, a spring, and a ring electromagnet. The collection box includes a cover plate detachably connected to the bottom of the collection box. The connecting seat is fixedly connected to the top of the collection box. The lower part of the suction pipe is fixedly connected to the top of the connecting seat. The connecting seat has a channel, and the lower part of the suction pipe has an opening communicating with the channel of the connecting seat. The top of the collection box has an inlet communicating with the channel of the connecting seat. The drainage pipe is connected to the rear of the suction pipe. The front support plate is fixedly connected to the rear end of the suction pipe. The left fixed rod is fixedly connected between the front and rear support plates, and the right fixed rod is fixedly connected between the front and rear support plates. The lead screw motor is fixedly connected to the rear support plate, and the sealing cover is connected to the end of the lead screw motor. A sealing ring is connected between the end of the cap and the end of the lead screw motor; the rear end of the cylinder is fixedly connected to the sealing cap, and the front end of the cylinder is fixedly connected to the front support plate. The front support plate has a central through hole, which communicates with the cylinder; the lead screw motor has a lead screw, the sealing cap has a central through hole, the lead screw passes through the central through hole of the sealing cap, the lead screw is located in the cylinder, the push shaft has an axial internal thread hole, the axial internal thread hole is connected and engaged with the lead screw, the push shaft passes through the central through hole of the front support plate, and the push shaft can move back and forth in the water suction pipe; the ferromagnetic metal disc is located in the water suction pipe, the edge of the ferromagnetic metal disc is close to the inner wall of the water suction pipe, the spring is located in the water suction pipe, the front end of the spring is fixedly connected to the ferromagnetic metal disc, and the rear end of the spring is fixedly connected to the front support plate; the annular electromagnet is fixedly connected to the outside of the water suction pipe, the annular electromagnet includes an annular iron core and a coil, and the coil is wound on the annular iron core.

2. The underwater suction device based on a spring-powered storage device according to claim 1, characterized in that, When the lead screw of the lead screw motor rotates, it moves the push shaft forward. The front end of the push shaft pushes the ferromagnetic metal disc forward to the position of the annular electromagnet. The ferromagnetic metal disc is located in front of the opening at the bottom of the water suction pipe. The spring is stretched, and the annular electromagnet attracts the ferromagnetic metal disc when energized. The lead screw rotates, causing the push shaft to move backward. The front end of the push shaft moves to the position of the front support plate. At this time, the marine organism capture system based on the spring storage device is in the initial state. In the initial state, the annular electromagnet is de-energized, and under the tension of the spring, the ferromagnetic metal disc moves backward to the position of the drain pipe.

3. The underwater suction device based on a spring-powered storage device according to claim 2, characterized in that, The drainage pipe includes a left drainage pipe and a right drainage pipe. The left drainage pipe is connected to the rear of the water intake pipe, and the right drainage pipe is connected to the rear of the water intake pipe.

4. The underwater suction device based on a spring-powered storage device according to claim 3, characterized in that, The left and right drainage pipes are arranged symmetrically.

5. The underwater suction device based on a spring-powered storage device according to claim 2, characterized in that, The spring is a spring that surrounds the push shaft.

6. The underwater suction device based on a spring-powered storage device according to claim 2, characterized in that, The number of springs is two or more, and the two or more springs are located around the push shaft, which can move back and forth in the space between the two or more springs.

7. The underwater suction device based on a spring-powered storage device according to claim 6, characterized in that, The spring includes a first spring, a fourth spring, a second spring, and a third spring.

8. A sea cucumber suction robot, comprising a main frame, characterized in that, The underwater suction device based on a spring-powered storage device, as described in any one of claims 1-7, is connected to the main frame.

9. A dynamic modeling method for an underwater suction device based on a spring-powered storage device as described in claim 7, characterized in that, Includes the following steps: A ferromagnetic metal disk is defined as the piston, a water suction pipe as the pipe, and the front support plate, rear support plate, left fixed rod, right fixed rod, screw motor housing, sealing cover, and cylinder as the support. A complete dynamic model including spring, piston, pipe, and support is constructed, and the physical information neural network PINN model is introduced to perform high-precision identification and joint solution of key fluid dynamic parameters. When the toroidal electromagnet is de-energized, the four pre-stretched springs simultaneously pull the piston, causing it to accelerate in the seawater. The axial dynamic equation is: In the formula, M is the piston mass, 4k is the total stiffness coefficient of the four springs, x0 is the initial tension of the springs, and x is the axial displacement of the piston relative to its initial position. The axial velocity of the piston; Where C is the piston cross-sectional area, D is the diameter of the suction pipe, and L is the length of the suction pipe, where L / D > 5; d ρ is the equivalent drag coefficient of the piston; ρ is the density of seawater. For fluid viscous resistance; reaction force F accel (t) and F pressure (t) is obtained from the output of the physical information neural network PINN, F pressure The value of (t) is equal to the suction force F at the suction working surface. suction (t); reaction force F accel (t) is: In the formula, L eff The equivalent acceleration length is given by α, where α is the velocity distribution correction factor, and v f (t) represents the average flow velocity inside the pipe; Suction F suction (t) is: The process of constructing the PINN (Physical Information Neural Network) model is as follows: Step (1), definition of physical governing equations and solution domain: Navier–Stokes equations: In the formula, u=(u r ,u θ ,u x ) represents the velocity vector, and p represents the pressure. Solution domain: includes the inlet region in front of the piston, the inside of the pipe, and the surrounding flow field region outside the pipe. The piston surface constitutes a time-varying motion boundary, while the inlet, outlet, and pipe wall are fixed boundaries. Step (2), Network Structure and Input / Output Design: Input quantities: spatial coordinates (x, r, θ), time t, and the axial displacement x(t) of the piston relative to its initial position and the axial velocity v of the piston. piston (t); Output quantities: Local pressure and velocity fields [p(x,r,θ,t),u] r ,u θ ,u x ], and a set of equivalent parameter vectors η to be identified: η=[α,C L ,L eff ,g] Step (3), the total loss function is defined as: L=λ1L PDE +λ2L BC +λ3L data Among them, the physical residual loss is in the internal sampling point set Ω int Above, definition: In the formula, The differential operator representing the Navier–Stokes equations; λ1, λ2, λ3 are weighting coefficients; Step (4), training and parameter identification process; Step S401: Obtain spatial coordinates (x, r, θ), time t, and axial displacement x of the piston relative to its initial position. (0) (t), piston axial velocity Step S402: In the corresponding spatiotemporal domain, internal points and boundary points are generated using Latin hypercube or random sampling to form a training set; Step S403, using spatial coordinates (x, r, θ), time t, and the axial displacement x of the piston relative to its initial position. (0) (t) and piston axial velocity As input, construct PINN and minimize the loss function L in the above equation to obtain the local pressure p, velocity field u, and equivalent parameter vector η; η=[α,C L ,L eff ,g] Step S404: Calculate each fluid force term using the identified parameters: Calculate the reaction force F accel (t): In the formula, the average flow velocity v in the pipe f (t) is: v f (t)≈v piston (t); Calculate the suction force F suction (t): Calculate the equivalent lateral force F lateral (t): Then, the piston motion is resolved using the following axial dynamic equations to obtain x. (1) (t), Next, the relative error of the displacement trajectory is calculated. If E exceeds the preset convergence threshold ∈ = 10 -3 Then determine x (1) (t) and the initial input x (0) If there is a significant difference in (t), then x will be... (1) (t), As new input, repeat steps S402-S404 for iterative training until E≤∈, that is, the flow field output by PINN converges with the motion state predicted by the dynamic equation.

10. The dynamic modeling method according to claim 9, characterized in that, The structure of the suction device was optimized: Step (S1): Manually define the spring stiffness k, pre-tension x0, piston mass M, suction pipe length L, and bracket installation length L. pipe The specific values, combined with spatial coordinates (x, r, θ), time t, axial displacement x(t) of the piston relative to its initial position, and axial velocity v of the piston, are determined by... piston (t), forming a set of data, then the spatial coordinates (x,r,θ), time t, axial displacement x(t) of the piston relative to the initial position, and axial velocity v of the piston are... piston (t) is input to the PINN model, and the PINN model outputs the local pressure p(x,r,θ,t) and the velocity field [u r ,u θ ,u x ],α,C L ,L eff ,ζ; Step (S2): Calculate the reaction force F using the following formula. accel (t): Calculate the suction force F using the following formula. suction (t): The effective lateral force F is calculated using the following formula. lateral (t): Step (S3): Calculate key performance indicators: The maximum equivalent stress σ of the stent is calculated using the following formula. max : In the formula, M bend (t)=F lateral (t)·L pipe c = D / 2 The maximum suction force F is calculated using the following formula. suction,max : F suction,max =max(F suction (t)) With the maximum equivalent stress σ of the stent max Under the constraint of not exceeding the allowable stress [σ] of the support material, the peak instantaneous suction force F is made to capture. suctin,max To maximize the target, the gradient descent method is used to find the optimal value using the automatic differentiation property of the PINN framework. The steps are as follows: Step 1), define the optimization objective function and construct the loss function J that includes the objective term and the penalty term. opt : J opt =-F suction,max +w·ReLU(σ max -[σ]) In the formula, the first term is the negative value of the maximum suction force; the second term is the penalty term for stress constraint, w is the penalty weight, and ReLU is the linear rectification function; Step 2), calculate the objective function J opt For design variables (k, x0, L, D, L) pipe gradient of ) Step 3), variable update: iteratively update the design variables along the negative gradient direction. In the formula, X=[k,x0,L,D,L pipe ] T The design variable vector is given, where γ is the learning rate; Step 4) Iteratively solve the problem by repeating the gradient calculation and variable update steps above until the design variables converge, thereby obtaining the optimal combination of structural parameters that satisfies the strength constraints and maximizes the suction force.

Images