Intelligent lifting support based on neural network

Abstract

The invention discloses an intelligent lifting support based on a neural network, and belongs to the field of intelligent manufacturing. The device comprises a base, a horizontal moving mechanism, a parallel lifting mechanism, a movable supporting plate and an image sensor. The horizontal moving mechanism realizes a horizontal moving function; the parallel lifting mechanism is connected with four connecting rods through an electric push rod to realize a parallel lifting function; the movable support plate is provided with a groove for positioning a workpiece; the visual sensor is installed at the bottom of the movable supporting plate, and the function of controlling the lifting position is achieved by collecting image information. During working, a workpiece is placed at the front end of the movable supporting plate, then the lifting plate is controlled to move through cooperation of the motor and the electric push rod, the fixed supporting plate is matched to achieve the workpiece conveying function, meanwhile, the visual sensor collects image information, and lifting is controlled through deep learning of a convolutional neural network model; the workpiece conveying device is exquisite in design and easy to operate, has certain universality, can position and convey workpieces, and facilitates follow-up machining of the workpieces in an assembly line.

Description

Technical Field

[0001] This invention relates to the field of machining, and more specifically to an intelligent lifting support based on neural networks. Background Technology

[0002] With the continuous development of machining technology, modern society is now capable of producing a variety of exquisite and complex workpieces. However, for complex or even irregular workpieces, current processing methods still have shortcomings. For example, the installation of some parts requires manual installation, which is time-consuming and labor-intensive, greatly reducing processing efficiency. Therefore, this invention is an intelligent lifting support based on neural networks. Combined with other devices, it can automatically realize the workpiece processing flow and transport the processed workpiece out. Furthermore, it uses sensors and a convolutional neural network model to precisely control the lifting function. Summary of the Invention

[0003] This invention discloses an intelligent lifting support based on a neural network, comprising a base, a horizontal moving mechanism, a parallel lifting mechanism, and a movable support plate. The horizontal moving mechanism includes a lifting plate I, a slider I, a slider II, a slider III, a slider IV, and a motor. The sliders connect the guide rail on the base to the lifting plate I, enabling horizontal movement. The parallel lifting mechanism includes a lifting plate II, an electric push rod, and connecting rods I, II, III, and IV. The four connecting rods connect the lifting plate I and the lifting plate II, allowing for relative movement. The electric push rod connects to the lifting plate II, enabling parallel lifting. The movable support plate is fixed to the lifting plate II and has grooves for workpiece positioning. During operation, the workpiece is first placed at the front end of the movable support plate, and then the movement of the lifting plate is controlled by the cooperation of the motor and the electric push rod, working in conjunction with the fixed support plate to transport the workpiece. This invention is ingeniously designed, simple to operate, and has a certain degree of versatility. It can position and transport workpieces and facilitate subsequent processing of workpieces in an assembly line.

[0004] The technical solution of this invention is: an intelligent lifting support based on neural networks, including a base 1, a horizontal moving mechanism 2, a parallel lifting mechanism 3, and a movable support 4; the base includes a base 1, a linear guide rail I6, a linear guide rail II14, and a motor 7 is installed on the top, with the linear guide rail I6 and the linear guide rail II14 each installed at both ends of the top; during operation, the lifting plate II13, under the combined action of the motor 7 and the electric push rod 10, raises the movable support plate 4 to the unloading position. After the workpiece is placed in, the movable support plate 4 first retracts to the unloading position, and then descends to position the workpiece on the fixed support plate. Repeating the above process can complete the function of workpiece transportation and processing at the same time.

[0005] The base 1 and linear guide rails I6 and II14 are all fixed together by bolts. Considering that heavy workpieces may be processed, the base 1 needs to have a large load capacity to support the entire device to complete the transportation and processing process.

[0006] The horizontal moving mechanism 2 includes a lifting plate I5, slider I8, slider II9, slider III15, slider IV16, motor 7, motor support, bearings, lead screw 18, and lead screw nut 19. Motor 7 is fixed to the motor support with screws, and the motor support is fixed to the top of the base 1 with bolts. The motor is connected to the lead screw 18 via bearings. The lead screw 18 is centrally located at the top of the base 1. Linear guide rails I6 and II14 are distributed on the left and right sides of the lead screw 18. Slider I8 and slider II9 are mounted on linear guide rail I6. Slider III15 and slider IV16 are mounted on linear guide II14. Lifting plate I5 is bolted to lead screw nut 19. The rotation of motor 7 drives lead screw 18 to rotate, which in turn drives lifting plate I5 to move linearly on linear guide I6 and linear guide II14 through lead screw nut 19. In order to control the movement range of horizontal moving mechanism 2 within 200mm, there is a protrusion in the middle of lifting plate I5. When lead screw nut 19 abuts against motor 7, it is the leftmost end of the movement, and when lead screw nut 19 abuts against the protrusion, it is the rightmost end of the movement.

[0007] The parallel lifting mechanism 3 includes an electric push rod 10, connecting rods I11, II12, III15, IV16, and a lifting plate II13. The electric push rod 10 is bolted to the lifting plate I5, and its end is bolted to the lifting plate II. Connecting rods I11, II12, III15, and IV16 are bolted to the lifting plates I5 and II13. The electric push rod 10 pushes the lifting plate II13 upwards, causing the four connecting rods to rotate, thus achieving parallel lifting of the lifting plate II13. When the electric push rod 10 is extended to its maximum length, the parallel lifting mechanism reaches its maximum movement, with a lifting height of 100mm.

[0008] The movable support plate 4 is fixed to the lifting plate II13 by bolts. The movable support plate 4 has a groove. The first section of the groove is shaped like the end of a single workpiece to be processed, and the remaining sections are shaped like the processed end. This groove serves to position the workpiece. The fixed support plate has the same shape as the movable support plate. The relative movement of the two support plates enables the transport of the workpiece.

[0009] The image sensor 21 is fixed to the base 1 via a connection, and has a data cable on its back for connecting to a computer, enabling it to capture and transmit data. In the traditional sense, an image sensor is a camera. A CCD is a semiconductor optoelectronic element manufactured using large-scale integrated circuit technology, and a CCD camera is a camera that utilizes this optoelectronic element. CCD cameras have excellent characteristics such as high resolution, low noise, low power consumption, and small size, and have quickly gained widespread application in industry.

[0010] Convolutional neural networks (CNNs) are deep learning models that have achieved tremendous success in the field of computer vision. Their design is inspired by the biological visual system and aims to mimic how humans process visual information. In recent years, CNNs have made significant progress in image recognition, object detection, image generation, and many other fields, becoming an important component of computer vision and deep learning research.

[0011] The beneficial effects of this invention are: the invention has a novel idea, simple structure, convenient operation, and strong adaptability. It can not only realize the positioning and transportation of workpieces to be transported, but also the positioning and transportation of completed workpieces. At the same time, other processing procedures of workpieces can be carried out simultaneously on this device. The shapes of the moving support plate and the fixed support plate can be flexibly adjusted as needed to realize the processing of different parts, which greatly saves labor costs and improves processing efficiency. Furthermore, this invention can accurately realize the function of the device through sensors and convolutional neural network models. Attached Figure Description

[0012] Figure 1 This is an isometric view of the entire invention;

[0013] Figure 2 This is a front view of the present invention;

[0014] Figure 3 This is the left view of the present invention;

[0015] Figure 4 This is a top view of the present invention;

[0016] Figure 5 This is an isometric view of the base of the present invention;

[0017] Figure 6 These are isometric views of lifting plate I and lifting plate II of the present invention;

[0018] Figure 7 This is a cross-sectional view of the slider of the present invention;

[0019] Figure 8 This is a cross-sectional view of the lead screw nut of the present invention;

[0020] Figure 9 This is an isometric drawing of the electric actuator of the present invention;

[0021] Figure 10 This is an isometric drawing of the connecting rod of the present invention;

[0022] Figure 11 This is an isometric drawing of the moving support plate of the present invention;

[0023] Figure 12 This is a schematic diagram of the movement of the movable support plate of the present invention;

[0024] Figure 13This is an isometric view of the image sensor of the present invention;

[0025] Figure 14 This is a schematic diagram of the convolutional neural network of this invention;

[0026] Figure 15 This invention relates to the imaging principle of the central perspective projection model.

[0027] Figure 16 This invention relates to the camera's commonly used coordinate system and imaging.

[0028] Figure 17 This is a schematic diagram illustrating the relationship between the camera coordinate system and the image coordinate system of this invention;

[0029] Figure 18 This is a schematic diagram of the two-dimensional chessboard calibration plate of the present invention;

[0030] Figure 19 This is a diagram of the convolutional neural network structure of the present invention;

[0031] Figure 20 This invention relates to the Sigmoid activation function code and image of the convolutional neural network model.

[0032] Figure 21 This is the code for the mean squared error loss function of the convolutional neural network model of this invention;

[0033] Figure 22 This is a flowchart of the process of this invention;

[0034] The labels in the diagram are as follows: 1-base, 2-horizontal moving mechanism, 3-parallel lifting mechanism, 4-moving support plate, 5-lifting plate I, 6-linear guide rail I, 7-motor, 8-slider I, 9-slider II, 10-electric push rod, 11-connecting rod I, 12-connecting rod II, 13-lifting plate II, 14-linear guide rail II, 15-slider III, 16-connecting rod III, 17-connecting rod IV, 18-lead screw, 19-lead screw nut, 20-slider IV, 21-image sensor. Detailed Implementation

[0035] The present invention will be further described below with reference to the accompanying drawings and embodiments, but the scope of the present invention is not limited to the description.

[0036] Example 1:

[0037] like Figure 2 As shown, a motor 7, linear guide rail I6, and linear guide rail II14 are fixed on the top of the base 1, and the two linear guide rails are respectively installed at both ends. The linear guide rails can move relative to the lifting plate I5 on them.

[0038] Furthermore, the horizontal moving mechanism 2 can be configured as follows: Figure 2 and Figure 3 The horizontal movement mechanism 2 shown includes a lifting plate I5, a lead screw 18, a lead screw nut 19, linear guide rails I6 and II14, a motor 7, sliders I8, II9, III15, and IV20. The lifting plate I5 is bolted to the lead screw nut 19. Four sliders are connected to the lifting plate I5 and mounted in pairs on the linear guide rails. When the motor 7 rotates, it drives the lead screw 18 to rotate, and the lead screw nut 19 on the lead screw 18 drives the entire lifting plate I5 to move horizontally, thus achieving the horizontal movement function. The base 1 also has a protrusion. When the lead screw nut 19 is against the dynamic torque sensor 23, it represents the leftmost end of the horizontal movement; when it is against the protrusion, it represents the rightmost end. The distance between the left end of the protrusion and the right end of the motor 7 is 200mm, thus controlling the horizontal movement range within 200mm.

[0039] Furthermore, the aforementioned parallel lifting mechanism 3 can be configured, such as... Figure 1 and Figure 3 The parallel lifting mechanism 3 shown includes a lifting plate II13, an electric push rod 10, connecting rods I11, II12, III16, and IV17. Both ends of the electric push rod 10 and the four connecting rods are bolted to the lifting plate I5 and lifting plate II13. This not only ensures the horizontal stability of the lifting plate II13 when stationary, but also achieves horizontal lifting through the movement of the electric push rod 10. When the electric push rod 10 lifts, since the lifting plate I5 remains stationary, it drives the four connecting rods to rotate, causing the lifting plate II13 to lift horizontally, thus achieving the parallel lifting function. When the push rod is at its longest extension, the parallel lifting mechanism 2 reaches its maximum lifting height of 100mm.

[0040] Furthermore, a movable support plate 4 can be provided, which is installed on the parallel lifting mechanism 3. The movable support plate 4 has grooves corresponding to the shape of the workpiece to achieve the function of workpiece positioning. The front section of the movable support plate 4 represents the shape of a single workpiece to be processed, and the remaining sections represent the shape of the workpiece after processing. The fixed support plate is designed the same as the movable support plate 4. Through the cooperation of the horizontal moving mechanism 2 and the parallel lifting mechanism 3, the transportation and installation of the workpiece can be completed simultaneously on the movable support plate 4 and the fixed support plate.

[0041] Furthermore, an image sensor 21 can be installed. The image sensor 21 is fixed to the base 1 via a connection, and a data cable on its back connects to a computer to achieve the function of capturing and transmitting data. In the traditional sense, an image sensor is a camera. A CCD is a semiconductor optoelectronic element manufactured using large-scale integrated circuit technology, and a CCD camera is a camera that utilizes this optoelectronic element. CCD cameras have excellent characteristics such as high resolution, low noise, low power consumption, and small size, and have quickly gained widespread application in industry.

[0042] The traditional visual imaging model is the central perspective projection model, also known as the pinhole imaging model. Mathematically, this model represents a central projection from three-dimensional space to a plane, described by a 3×4 projection matrix M. The central perspective projection model assumes that all light reflected or emitted from the object's surface passes through a pinhole and is projected onto the image plane. This pinhole is called the optical center or projection center. Figure 15 This describes the imaging principle of a central perspective projection model. The model mainly consists of an optical center, an image plane, and an optical axis. The image distance *v* from the optical center to the image plane is called the focal length *f*, and the object distance *u* is equal to the distance from the optical center to the object being measured. The object point *P*, the optical center *C*, and the corresponding image point *p* are all on a straight line.

[0043] According to the imaging principle of central perspective projection, the distance X from object point P to the optical axis and the distance x from the corresponding image point p to the optical axis satisfy the proportional relationship between corresponding sides of similar triangles in geometric optics, i.e.

[0044]

[0045] In machine vision applications, there is a certain relationship between the 3D coordinates of a point on the surface of a spatial object and the coordinates of its corresponding point in the image. This relationship can be determined by establishing a geometric model of camera imaging; the parameters of the geometric model are the camera parameters. Therefore, it is necessary to first establish commonly used camera coordinate systems, including the camera coordinate system, image coordinate system, pixel coordinate system, and world coordinate system. The camera coordinate system, image coordinate system, and pixel coordinate system are as follows: Figure 16 As shown.

[0046] O, the origin of the camera coordinate system c That is, the optical center of the camera, z c The axis coincides with the camera's optical axis, and the imaging direction is taken as positive. c axis, y c The axes are parallel to the x-axis and y-axis of the image coordinate system, respectively. To describe the coordinates of a pixel in the image, the top-left vertex O of the imaging plane is used. i A pixel coordinate system is established with the origin as the origin, and the u-axis and v-axis are parallel to the x-axis and y-axis of the image coordinate system, respectively. Each image is stored as an m×n array sequence, where each element is a pixel. The pixel coordinates (u, v) represent the row and column numbers of that image point in the image plane. The pixel coordinate system is measured in pixels. The world coordinate system, also known as the global coordinate system, is typically considered as a whole, taking the measured object and the camera as a single unit. The position of a point P in space is usually described by its coordinates (x, y, z) in the world coordinate system.

[0047] (1) Relationship between camera coordinate system and world coordinate system:

[0048] The relative relationship between any two coordinate systems can be decomposed into a rotation around the origin and a translation. Rotation can be expressed in various ways, such as Euler angles, rotation vectors, and quaternions. Below, we use Euler angles to represent the rotation of the coordinate system. Let the object point P(x,y,z) have coordinates (x...y...z) in the camera coordinate system. c ,y c ,z c Then (x) can be described using a rotation matrix and a translation vector. c ,y c z c The relationship between (x,y,z) and (x,y,z):

[0049] Equation (5.2) can be rewritten as

[0050]

[0051] Here, R is called the rotation matrix, which is a 3×3 identity orthogonal matrix, where the elements r0 to r8 are the rotation angles (α). x ,α y ,α z The trigonometric function combination of α, the rotation angle (α) x ,α y ,α z Euclidean angles are defined as the Euler angles obtained by transforming the world coordinate system to be consistent with the camera coordinate system's orientation, rotating the system around each of the three coordinate axes. T = [T x T y T z ] T This is called the translation vector, and it is the origin O of the world coordinate system. w The coordinates in the camera coordinate system, which is the coordinates of the world coordinate system origin O. w Move to the origin O of the camera coordinate system c The translation amount. Through rotation and translation, the world coordinate system and the camera coordinate system are made to coincide.

[0052] Let the world coordinate system revolve around x w Axis rotation α x The resulting rotation matrix is ​​R x , around y w Axis rotation α y The resulting rotation matrix is ​​R y , around z w Axis rotation α z The resulting rotation matrix is ​​R z According to the coordinate transformation relationship, R x R y R z They are respectively

[0053]

[0054] For two coordinate systems whose relative relationship is determined, the values ​​of each element of the rotation matrix R and translation vector T are also determined. However, if the world coordinate system is specified to rotate around each coordinate axis in different rotation orders, different rotation angle values ​​and rotation matrix expressions will be obtained. For example, in photogrammetry, it is common practice to have the world coordinate system rotate around the y-axis first. w Axis rotation a y Then around the current x w Axis rotation a x Finally, around the current z w Axis rotation a z Then the rotation matrix R is

[0055] R = R z R x R y (1.5)

[0056] It is easy to verify that the rotation matrix R is an orthogonal identity matrix. In the commutative equation (1.5), R... x R y R z By determining the order of rotations, we can obtain expressions for the rotation matrix R under different rotation orders.

[0057] (2) Relationship between camera coordinate system and image coordinate system:

[0058] like Figure 17 As shown, the relationship between the camera coordinate system and the image coordinate system can be derived from the basic relationship of the central perspective projection model [Equation (5.1)]. The image coordinates (x, y) of image point p and the camera coordinates (x, y) of object point P are related. c ,y c ,z c The relationship is

[0059]

[0060] Where f is the focal length. Equation (1.6) can be written in homogeneous coordinate form as follows:

[0061]

[0062] (3) Relationship between pixel coordinate system and image coordinate system:

[0063] Assuming the coordinates of the center pixel of the image are (u0, v0), and the physical dimensions of each pixel of the camera's image sensor in the x and y directions are d... x d y Therefore, the relationship between the coordinates (x, y) in the image coordinate system and the coordinates (u, v) in the pixel coordinate system can be expressed as:

[0064] Written in matrix form as

[0065] Written in homogeneous coordinate form as

[0066]

[0067] The above describes the case where both the image coordinate system and the pixel coordinate system are Cartesian coordinate systems. However, in most cases, the angle between the two axes of the pixel coordinate system is θ.

[0068] Written in homogeneous coordinate form as

[0069]

[0070] (4) Relationship between pixel coordinate system and world coordinate system:

[0071] By obtaining the transformation relationships between the various coordinate systems through the previous steps, we can further derive the transformation relationship between pixel coordinates and the world coordinate system:

[0072]

[0073] Define a 3×4 matrix M:

[0074]

[0075] Among them, f x =f / d x f y =f / d y , are called the normalized focal lengths on the x-axis and y-axis, respectively. (u0, v0) and normalized focal length (f) x ,f y The translation vector T, rotation angle, and rotation matrix R are the camera's intrinsic parameters, describing the camera's inherent characteristics. The translation vector T, rotation angle, and rotation matrix R are the camera's extrinsic parameters, describing the relative position and attitude between the camera coordinate system and the world coordinate system. In the two matrices constituting matrix M (i.e., the two matrices on the right-hand side of the equation), the first matrix consists of the camera's intrinsic parameters and is called the intrinsic parameter matrix; the second matrix consists of the camera's extrinsic parameters and is called the extrinsic parameter matrix. The central perspective projection imaging relationship can be described by matrix M:

[0076]

[0077] Matrix M describes the central perspective projection relationship from a point in space to a pixel, and is called the projection matrix. Expanding matrix M yields the elements of the projection matrix:

[0078]

[0079] Due to Z cZ is the projection of the distance from object point P to image center C along the optical axis, therefore Z c ≠0. Expanding equation (1.15), we obtain the collinearity equation described by the elements of the projection matrix:

[0080]

[0081] The collinearity equation and projection matrix in the central perspective projection model are the most fundamental and important relationships in photogrammetry. Almost all photogrammetric theories start from this point and are based on it.

[0082] The central perspective projection imaging relationship can be described by matrix M. Vision system calibration devices based on calibration boards are among the most widely used camera calibration devices. Zhang Zhengyou et al. simplified the three-dimensional calibration block into a two-dimensional checkerboard calibration board. Figure 18 This greatly reduces the processing requirements of the calibration material without sacrificing its calibration accuracy.

[0083] The camera model [Equation (1.13)] can be rewritten as follows:

[0084]

[0085] The intrinsic parameter matrix is ​​described by matrix K.

[0086]

[0087] The camera model can then be rewritten as:

[0088]

[0089] Assuming the world coordinate system plane coincides with the plane containing the calibration plate, i.e., z = 0, then equation (1.20) can be transformed into:

[0090]

[0091] Among them, R i (i = 1, 2, 3) represents the i-th column vector of the rotation matrix R. The homography matrix H is defined as:

[0092]

[0093] The intrinsic parameter matrix K can then be solved using the following constraints:

[0094] [h1 h2 h3]=λK[R1 R2 T] (1.23)

[0095] Where λ is an arbitrary scalar.

[0096] Since R is an orthogonal matrix, it has the following properties:

[0097]

[0098] Using equations (1.23) to (1.25), we can obtain:

[0099]

[0100] h1 and h2 are obtained by solving the homography matrix H, so the only unknown is matrix K. Taking three photos of different calibration planes will yield three different homography matrices H. Under two constraints, six equations can be generated, which can then be used to solve for the five unknowns in matrix K.

[0101] Define the symmetric matrix B as:

[0102]

[0103]

[0104] The extrinsic parameter matrix can be solved using the intrinsic parameter matrix:

[0105]

[0106] Furthermore, convolutional neural networks (CNNs) can be set up, a type of deep learning model that has achieved great success in the field of computer vision. They are inspired by the biological visual system and designed to mimic how humans process vision. In the past few years, CNNs have made significant progress in image recognition, object detection, image generation, and many other fields, becoming an important part of computer vision and deep learning research.

[0107] Figure 19 This is a simple convolutional neural network structure diagram. The first layer takes an image as input, performs a convolution operation, and obtains a second layer with a depth of 3 feature maps. Pooling is then performed on the second layer feature maps to obtain a third layer feature map with a depth of 3. This process is repeated to obtain a fifth layer feature map with a depth of 5. Finally, these five feature maps, i.e., five matrices, are unfolded row-wise and concatenated into a vector, which is then fed into a fully connected layer, which is a backpropagation (BP) neural network. Each feature map in the diagram can be viewed as a neuron arranged in a matrix, very similar to the neurons in a BP neural network. The calculation process for convolution and pooling is shown below.

[0108] For an input image, it is transformed into a matrix, where each element represents a pixel value. For example, a 5×5 image can be convolved using a 3×3 kernel to obtain a 3×3 feature map. The convolution kernel is also called a filter.

[0109] Generally, the input image matrix, the subsequent convolution kernel, and the feature map matrix are all square matrices. Here, let the input matrix size be w, the convolution kernel size be k, the stride be s, and the number of zero-padding layers be p. Then, the formula for calculating the size of the feature map generated after convolution is:

[0110]

[0111] Pooling, also known as downsampling, is the opposite of upsampling. Features obtained from convolution... Figure 1 A pooling layer is generally needed to reduce the amount of data.

[0112] After several layers of convolution and pooling operations, the resulting feature maps are unfolded row by row, concatenated into vectors, and then input into a fully connected network.

[0113] enter:

[0114] V = conv2(W, X,"valid") + b (1.31) Output:

[0115]

[0116] The input-output formulas above apply to each convolutional layer, which has a different weight matrix W, and W, X, and Y are in matrix form. For the last fully connected layer, let's call it the L-th layer, the output is a vector y. L If the expected output is d, then the total error formula is:

[0117] Total error:

[0118]

[0119] `conv2()` is a convolution function in Matlab. The third parameter, `valid`, specifies the type of convolution operation. `W` is the convolution kernel matrix, `X` is the input matrix, and `b` is the bias. It is the activation function. In the total error, d and y are the vectors of the expected output and the network output, respectively. ||x||2 represents the 2-norm of vector x, calculated as follows: The input and output calculation formulas for neurons in fully connected layers are exactly the same as those for backpropagation (BP) networks.

[0120] Gradient formulas for convolutional and pooling layers:

[0121]

[0122] A convolutional neural network (CNN) is a feedforward neural network where each neuron is connected only to the neurons in the preceding layer, receives the output of the previous layer, performs calculations, and then outputs the result to the next layer. There is no feedback between layers. Therefore, to calculate local gradients, the gradients of neurons in the next layer can be recursively passed forward.

[0123] The recurrence relation for δ is obtained as follows:

[0124]

[0125] The activation function and loss function of this convolutional neural network model are as follows.

[0126] Sigmoid function:

[0127]

[0128] Code and images such as Figure 20 As shown.

[0129] Mean squared error loss function:

[0130]

[0131] Code as follows Figure 21 As shown.

[0132] The working principle of this invention is:

[0133] like Figure 11 As shown, before operation, the movable support plate 4 is placed at the feeding station. The robot arm places a single workpiece on the first section of the movable support plate 4. During operation, the movable support plate 4 retracts to the unloading station via the horizontal moving mechanism 2, and then descends via the cooperation of the horizontal moving mechanism 2 and the parallel lifting mechanism 3 to position the workpiece on the fixed support plate. Subsequently, it moves forward to the feeding position via the horizontal moving mechanism 2, and then rises to the unloading station via the cooperation of the horizontal moving mechanism 2 and the parallel lifting mechanism 3 to place the second workpiece. After the above process, the two workpieces are transported to the next station to complete the next processing operation. Finally, the process is completed from a single unprocessed workpiece to multiple processed workpieces. Finally, the movable support plate 4 transports the workpieces to the roller conveyor mechanism for transport.

[0134] The horizontal moving mechanism 2 consists of a lead screw nut 19 and sliders mounted on the base. The lead screw nut 19 controls the horizontal movement of the lifting plate I5, which in turn drives the upper movable support plate 4 to move horizontally. Four sliders are mounted on the four corners of the lifting plate I5 and on the guide rails at both ends of the base 1 to ensure that the lifting plate I5 remains horizontal when moving.

[0135] The parallel lifting mechanism 3 consists of an electric push rod 10 and a connecting rod connecting the lifting plate I5 and the lifting plate II13. By driving the electric push rod 10, the lifting plate II13 is lifted and the four push rods are rotated, thus achieving the parallel lifting of the lifting plate I13. Due to the tilt angle of the push rods and the rotation of the connecting rods, horizontal displacement will occur during parallel lifting. Therefore, it is necessary to consider the motor of the horizontal movement mechanism and the electric push rod 10 of the parallel lifting mechanism 2, and control the movement range of the horizontal lifting mechanism 2 within 100mm. The tilt angle of the push rods is determined. With the cooperation of the motor 7 and the electric push rod 10, the lifting of the movable support plate can be achieved.

[0136] The structures of the lifting plate I5 and the lifting plate II13 are as follows: Figure 5 As shown, the four lifting lugs are connected to four connecting rods during installation, which in turn connect the horizontal moving mechanism 2 and the parallel lifting mechanism 3. The structure of the electric push rod 10 is as follows: Figure 8 As shown, both ends are fixed to lifting plate I5 and lifting plate II13 respectively by bolts, realizing the function of lifting plate II. The linkage structure is as follows. Figure 9 As shown, the two ends of the four connecting rods are respectively connected to the lifting plate I5 and the lifting plate II13, and together with the electric push rod 10, they realize the function of parallel lifting.

[0137] The movable support plate structure is as follows: Figure 10 As shown, the upper end has a groove to position the workpiece. The movable support plate is installed on the lifting plate II13. During installation, it can be appropriately raised as needed. The movable support plate achieves its position through the combined action of the horizontal moving mechanism 2 and the parallel lifting mechanism 3. Figure 11 The function shown enables the workpiece to be transported and installed on the moving support plate 4 and the fixed support plate.

[0138] The image sensor, such as Figure 13 As shown, it is installed on the base to capture image information and convert it into digital signals for transmission to the computer.

[0139] The convolutional neural network is a deep learning model that has achieved great success in the field of computer vision. This invention uses this model to achieve precise execution of the lifting device's functions through its powerful image recognition capabilities.

[0140] Matters not covered in this invention are common knowledge.

[0141] The above embodiments are only for illustrating the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They should not be construed as limiting the scope of protection of the present invention. All equivalent changes or modifications made in accordance with the spirit and essence of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A smart lifting support based on neural networks, characterized in that: The system includes a base (1), a horizontal moving mechanism (2), a parallel lifting mechanism (3), and a movable support plate (4). The base (1) includes a linear guide rail I (6) and a linear guide rail II (14). The linear guide rail I (6), the linear guide rail II (14), and the lead screw nut (19) in the horizontal moving mechanism (2) are connected to the lifting plate I (5) in the horizontal moving mechanism (2), allowing for relative movement. The horizontal moving mechanism (2) is driven by a motor (7) to drive the lead screw nut (19) on the lead screw (18) to move the lifting plate I (5), thereby achieving the horizontal moving function. The lifting plate I (5) and the lifting plate II (13) in the parallel lifting mechanism (3) are connected by four connecting rods and can move relative to each other; the parallel lifting mechanism (3) lifts the lifting plate II (13) by an electric push rod (10) to achieve the parallel lifting function; the lifting bracket has an image sensor (21) to monitor and collect the relative position data between the lifting plate I (5) and the lifting plate II (13) and transmits the data to the computer through a data cable; the computer has a built-in trained convolutional neural network model, which achieves precise control of the lifting position of the lifting bracket through deep learning.

2. The intelligent lifting support based on a neural network according to claim 1, characterized in that: The horizontal moving mechanism (2) includes a lifting plate I (5), slider I (8), slider II (9), slider III (15), slider IV (20), motor (7), motor support, bearing, lead screw (18), and lead screw nut (19); the motor (7) is fixed to the motor support by screws, the motor support is fixed to the top of the base (1) by bolts, the motor is connected to the lead screw (18) through the bearing, the lead screw (18) is centrally arranged at the top of the base (1), and the linear guide I (6) and the linear guide... Guide rail II (14) is distributed on the left and right sides of the lead screw (18). Slider I (8) and slider II (9) are installed on linear guide rail I (6). Slider III (15) and slider IV (20) are installed on linear guide rail II (14). Lifting plate I (5) is connected to lead screw nut (19) by bolts. The motor (7) rotates and drives lead screw (18) to rotate, and then drives lifting plate I (5) to move linearly on linear guide rail I (6) and linear guide rail II (14) through lead screw nut (19).

3. The intelligent lifting support based on a neural network according to claim 1, characterized in that: The parallel lifting mechanism (3) includes an electric push rod (10), connecting rod I (11), connecting rod II (12), connecting rod III (16), connecting rod IV (17), and lifting plate II (13). The electric push rod (10) is fixed to the lifting plate I (5) by bolts, and the end of the push rod is fixed to the lifting plate II (13) by bolts. Connecting rod I (11), connecting rod II (12), connecting rod III (13), and connecting rod IV (14) are fixed to the lifting plate I (5) and the lifting plate II (13) by bolts. The electric push rod (10) pushes and drives the four connecting rods to rotate, thereby realizing the parallel lifting of the lifting plate II (13).

4. A lifting support mechanism based on a convolutional neural network according to claim 1, characterized in that: The tilt angle of the electric push rod (10) is designed based on the work requirements. Through the cooperation of the motor (7) and the electric push rod (10), the track is transported on the moving support plate (4).

5. A neural network-based intelligent lifting support according to claim 1, characterized in that: The groove on the movable support plate (4) can be flexibly changed according to the shape of the workpiece being processed, so as to realize the positioning and transportation of workpieces of different shapes.

6. The intelligent lifting support based on a neural network according to claim 1, characterized in that: The image sensor (21) is a CDD camera image sensor, which is mounted on the base (1). It captures the light above, converts the light into electrical charge, converts it into a digital signal through an analog-to-digital converter chip, and transmits the digital signal to the computer.

Images