High speed collating mechanism for inductor coils

Abstract

The application discloses a high-speed arrangement mechanism for an inductor coil, which comprises a rack, at least one vibrating feeding unit arranged on the rack, each vibrating feeding unit comprising a vibrating feeder and a tray, the tray being arranged at the outlet of the vibrating feeder and being used for laying the inductor coil by vibration, a visual identification system comprising an industrial camera arranged directly above the tray, the industrial camera being configured to collect images of the inductor coil laid on the tray, a mechanical hand unit comprising a translatable driving module arranged on the rack and a suction disc assembly driven by the driving module, the suction disc assembly comprising at least one independently controlled suction disc unit and being capable of independently performing vertical extension and contraction and rotation around its own axis, and a coil arranger arranged on the rack and used for receiving and arranging the inductor coils with uniform directions.

Description

Technical Field

[0001] This invention relates to the field of inductor coil technology, and more particularly to a high-speed inductor coil straightening mechanism. Background Technology

[0002] In the field of electronic component manufacturing and assembly, inductors are a fundamental component. Subsequent processing, such as lead soldering, performance testing, and mounting, typically requires the coils to be arranged and positioned in a specific orientation. Traditional production relies heavily on manual handling of material preparation and orientation correction, resulting in low efficiency and poor consistency. Although automated feeding systems have emerged, these systems only achieve rough feeding, failing to ensure that each inductor coil is laid out flat and without overlap, affecting the success rate of subsequent picking. Furthermore, existing feeding systems have significant positional deviations, still requiring manual intervention or secondary adjustments, making it difficult to meet the requirements of high-speed, high-precision production. Summary of the Invention

[0003] To address at least one of the aforementioned technical problems, the present invention provides a high-speed coil straightening mechanism.

[0004] This invention provides a high-speed inductor coil conditioning mechanism, comprising:

[0005] frame;

[0006] At least one vibratory feeding unit is disposed on the frame; each vibratory feeding unit includes a vibratory feeder and a material tray; the material tray is disposed at the outlet of the vibratory feeder and is configured to lay out the inductor coil by vibration;

[0007] A visual recognition system includes an industrial camera positioned directly above the tray, the industrial camera being configured to capture images of inductor coils laid flat on the tray;

[0008] The robotic arm unit includes a translational drive module mounted on the frame and a suction cup assembly driven by the drive module; the suction cup assembly includes at least one independently controlled suction cup unit, each of which can independently perform vertical telescopic movement and rotational movement around its own axis.

[0009] A coil arranger, mounted on the frame, is used to receive and arrange inductor coils with uniform orientation;

[0010] The visual recognition system is communicatively connected to the robotic arm unit, used to identify the orientation and coordinates of each inductor coil, and to control the robotic arm unit to grasp and correct the inductor coil before placing it in the coil arranger.

[0011] Preferably, the visual recognition system has a built-in image processing module, which executes the following algorithm to identify the positive and negative directions and center coordinates of the inductor coil, including:

[0012] The industrial camera is controlled to capture images of the material tray, and Gaussian filtering is used to denoise the images;

[0013] Based on the denoised image, edge detection is performed using the Canny operator, and the circular outlines of all inductor coils in the image are identified using the Hough circle transform algorithm. The center coordinates of the circular outlines are calculated as the center coordinates of the inductor coils, and the first region image of each coil is located.

[0014] Based on the first region image of each location, a local adaptive binarization method based on gray value variance is used to process it and generate a binarized image.

[0015] Extract the largest connected component from the binarized image and label it as the fin region image; calculate the minimum bounding rectangle of the fin region image, and define the direction of the long side of the minimum bounding rectangle as the orientation of the inductor coil.

[0016] Preferably, the image processing module is further configured to perform the following steps to verify and correct the recognition results, including:

[0017] Pre-collect inductor coil image samples including upright, reverse, tilted, overlapping, and various lighting conditions to construct a training dataset; use the training dataset to train a convolutional neural network to obtain a classification verification model, the classification verification model is configured to take the cropped coil image as input and output the coil orientation classification result and confidence level;

[0018] The first region image of each coil is input into the classification verification model, and the coil orientation classification result and confidence score are output.

[0019] When the orientation of the output coil is consistent with the orientation determined according to the minimum bounding rectangle, the output result of the classification verification model is taken as the final orientation.

[0020] When the orientation of the output coil is inconsistent with the orientation determined by the minimum bounding rectangle, it is determined whether the confidence level is higher than the preset confidence threshold. If yes, the output result of the classification verification model is used as the final orientation. If no, the orientation determined by the minimum bounding rectangle is maintained as the final orientation, and the corresponding coil is marked as requiring manual re-inspection.

[0021] Preferably, there are two vibratory feeding units, symmetrically arranged on the left and right sides of the frame, and the coil arranger is arranged between the two vibratory feeding units.

[0022] Preferably, the suction cup assembly includes four independently controlled suction cup units, forming a four-joint suction cup assembly.

[0023] Preferably, the rotation angle of each suction cup unit in the four-joint suction cup assembly is independently controlled by the coil orientation identified by the visual recognition system.

[0024] Preferably, a piezoelectric ceramic vibrator is installed at the bottom of the material tray, and the piezoelectric ceramic vibrator is connected to the controller;

[0025] The controller is configured to first drive the vibrator at a first frequency to concentrate the inductor coils, and then drive the vibrator at a second frequency higher than the first frequency to spread the inductor coils out.

[0026] Preferably, the coil arranger is provided with a plurality of positioning slots that match the gripping layout of the robotic arm unit, and the positioning slots include coil center positioning pins and coil fin limiters.

[0027] Preferably, the discharge track of the vibrating feeder is provided with a guide member, the width of which is greater than the diameter of a single inductor coil and less than the sum of the diameters of two inductor coils, so as to control the inductor coils to be output in a single row in an orderly manner.

[0028] Preferably, the frame is provided with a material picking end at the end of the coil arranger. When the positioning slot on the coil arranger is filled, the arranged inductor coil is sent to the material picking end and waits to be transported to the next process.

[0029] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0030] This invention provides a high-speed inductor coil sorting mechanism, comprising a frame; at least one vibratory feeding unit disposed on the frame; each vibratory feeding unit comprising a vibratory unloader and a tray; the tray being disposed at the outlet of the vibratory unloader and configured to lay inductor coils flat by vibration; a vision recognition system comprising an industrial camera disposed directly above the tray, the industrial camera being configured to capture images of the inductor coils laid flat on the tray; a robotic arm unit comprising a translational drive module disposed on the frame, and a suction cup assembly driven by the drive module; the suction cup assembly comprising at least one independently controlled suction cup unit, each suction cup unit being capable of independent vertical extension and retraction and rotation around its own axis; and a coil sorter disposed on the frame for receiving and arranging inductor coils with uniform orientation; wherein the vision recognition system is communicatively connected to the robotic arm unit for identifying the orientation and coordinates of each inductor coil, and controlling the robotic arm unit to grasp and correct the inductor coils before placing them in the coil sorter.

[0031] This invention achieves initial flattening of inductor coils through a vibratory feeding unit. A visual recognition system acquires real-time image information of the coils, identifying their position and orientation. The system then controls a multi-suction cup assembly in the robotic arm unit to independently grasp, rotate, and vertically lift the coils, enabling precise grasping and active orientation correction for each inductor coil. Finally, the coils with uniform orientation are neatly arranged in a coil arranger, significantly improving material handling efficiency and positioning accuracy, reducing manual intervention, and meeting the needs of modern electronic assembly lines for high-speed, high-consistency production.

[0032] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention or the background art, the accompanying drawings used in the embodiments of the present invention or the background art will be described below.

[0034] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the specification, serve to illustrate the technical solutions of this disclosure.

[0035] Figure 1 This is a schematic diagram of the structure of a high-speed inductor coil tidying mechanism provided in an embodiment of the present invention;

[0036] Figure 2 A schematic diagram illustrating the process of a visual recognition system for identifying the positive and negative directions and center coordinates of an inductor coil, as provided in an embodiment of the present invention;

[0037] Figure 3 This is a schematic diagram of the structure of a high-speed inductor coil tidying mechanism provided in another embodiment of the present invention;

[0038] Figure 4 This is a partially enlarged structural diagram of the coil arranger 50 provided in an embodiment of the present invention. Detailed Implementation

[0039] To enable those skilled in the art to better understand the present invention, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0041] Please see Figure 1 , Figure 1 This is a schematic diagram of a high-speed inductor coil straightening mechanism provided in an embodiment of the present invention. Figure 1 As shown, the high-speed coil straightening mechanism includes:

[0042] Rack 10;

[0043] At least one vibratory feeding unit 20 is disposed on the frame 10; each vibratory feeding unit 20 includes a vibratory feeder 201 and a material tray 202; the material tray 202 is disposed at the outlet of the vibratory feeder 201 and is configured to lay out the inductor coil by vibration.

[0044] The visual recognition system 30 includes an industrial camera positioned directly above the tray 202, the industrial camera being configured to capture images of inductor coils laid flat on the tray 202;

[0045] The robotic arm unit 40 includes a translational drive module 401 mounted on the frame 10, and a suction cup assembly 402 driven by the drive module 401; the suction cup assembly 402 includes at least one independently controlled suction cup unit 4020, each of the suction cup units 4020 being capable of independently performing vertical telescopic movement and rotational movement around its own axis.

[0046] A coil arranger 50 is disposed on the frame 10 and is used to receive and arrange inductor coils with uniform orientation.

[0047] The visual recognition system 30 is communicatively connected to the robotic arm unit 40, and is used to identify the orientation and coordinates of each inductor coil, and control the robotic arm unit 40 to grasp and correct the inductor coil before placing it in the coil arranger 50.

[0048] In this embodiment, the vibratory feeding unit 20 is responsible for initially sorting, orienting, and individually and smoothly conveying the randomly stacked inductor coils to the visual recognition station. Typically, the vibratory feeding unit 20 can consist of a vibratory feeder 201 and a tray 202. The core component of the vibratory feeder 201 is a hopper with an internal spiral ascending track. The hopper wall is equipped with an electromagnet or piezoelectric ceramic drive device, which generates frequency vibration through pulse vibration, causing the coils inside the hopper to creep, flip, and advance along the track. The track is usually equipped with screening and guiding mechanisms. For example, baffles / gaps allow only coils with a specific orientation (e.g., fins facing upwards) to pass through; coils that do not meet the requirements are scraped back to the bottom of the hopper for re-screening. Gaps are also needed to allow coils that are too small or overlapped to fall through. The tray 202 is connected to a flat vibratory track below the outlet of the vibratory feeder 201. It receives the coils from the hopper and disperses and flattens them through micro-vibration, preventing stacking. The vibratory feeding unit 20 can automatically complete the initial sorting and conveying of large batches of messy coils, significantly reducing the labor intensity and time cost of manual loading and sorting. Through the ingenious track design (such as stops and gaps) inside the vibratory unloader 201, coils with the required posture (such as fins facing upwards) can be effectively screened out, and overlapping or stuck coils are removed and returned to the hopper, providing preprocessing for subsequent visual recognition and improving the overall recognition success rate. The secondary vibration of the tray 202 spreads the coils evenly, avoiding stacking and obstruction, and ensuring that each coil can be clearly captured by the industrial camera, which is a prerequisite for achieving high-precision and high-reliability recognition.

[0049] Traditional feeding systems often fail to identify the position and orientation of coils, affecting sorting efficiency and the final result. Therefore, this embodiment incorporates a vision recognition system 30 in its sorting mechanism. Its core task is to quickly and accurately identify the center coordinates and rotation angle (i.e., orientation) of each coil. Typically, the vision recognition system 30 includes an industrial camera, a light source, and image processing software. The industrial camera is usually a high-frame-rate area-array CMOS camera or a line-array camera. It is mounted on a fixed frame directly above the material tray 202, ensuring the field of view covers the entire working area of ​​the tray 202. The light source typically uses a ring-shaped shadowless lamp or a strip of combined light, illuminating from around the camera or at a specific angle. The purpose of the lighting is to highlight the height difference or contour difference between the fins and the coil body; for example, low-angle lighting creates a distinct shadow on the fins, significantly reducing the difficulty of image processing. The industrial camera can be either a 2D or 3D camera, directly acquiring height information for more direct fin identification. Therefore, the vision recognition system 30 can quickly and accurately identify the center coordinates of each coil, providing millimeter-level positioning guidance for the robotic arm's gripping, ensuring the accuracy and reliability of the gripping process. It can accurately identify the coil orientation (fin direction), solving the problems of fatigue and errors that are prone to occur during manual visual inspection, and ensuring the consistency of product arrangement direction.

[0050] The robotic arm unit 40 is responsible for performing the physical actions of grasping, aligning, and placing. The drive module 401 is used to realize the movement along the X and Y axes in the horizontal plane. It typically employs a gantry or cantilever structure. The gantry structure offers the best rigidity, stable operation, and high precision, making it the preferred choice for high-speed, high-precision applications. The cantilever structure is simple and low-cost, suitable for scenarios with small loads and low cycle time requirements. The drive element is typically a servo motor + precision ball screw or linear motor. The suction cup assembly 402 typically includes a suction cup unit 4020, an independent rotating mechanism, and an independent telescopic mechanism. The suction cup unit 4020 typically uses a vacuum suction cup, controlled by a solenoid valve. The suction cup material and shape must be selected based on the flatness and weight of the coil surface to ensure stable grasping without damaging the product. The independent rotating mechanism, i.e., a small servo motor or rotary cylinder connected below each suction cup, drives the suction cup to rotate, thereby enabling real-time alignment of the coil in mid-air. The independent telescopic mechanism allows each suction cup to be equipped with a small cylinder or electric push rod, enabling independent up-and-down movement to accommodate coils of different heights or with slight undulations. The suction cup assembly 402 follows instructions from the vision system. After receiving coil coordinate and angle information, the motion controller plans the optimal path, controlling the drive module 401 to move above the target, lower the suction cup, pick up the coil, and then synchronously rotate to a uniform angle during the lifting process. Finally, it moves above the coil arranger 50 and lowers the coil.

[0051] Therefore, in the robotic arm unit 40 of this embodiment, each suction cup can rotate immediately after gripping, uniformly correcting coils with different orientations to the target direction. This eliminates the step of adjusting the posture on the material tray 202 before gripping, greatly shortening the cycle time and achieving true high-speed sorting. The suction cup assembly 402 can grip multiple coils simultaneously and complete the correction and placement of multiple coils in one stroke, significantly improving production efficiency compared to a single suction cup device.

[0052] The coil arranger 50 is responsible for receiving and orderly arranging coils that have been aligned to the same direction, and then outputting them as a tray. It is typically a fixture or tray with precision positioning slots 501, such as... Figure 3As shown. The shape of the slot matches the shape of the inductor coil, ensuring that the coil is fixed in position after being placed. It is mounted on a linear module and can move in steps in the horizontal or vertical direction. After each set of coils is placed, the sorter moves one station to make room for the next set of coils in the positioning slot 501. Preferably, a rotary tray 202 can be used to switch stations by rotation, or a conveyor belt can be used to continuously place multiple fixtures to complete the coil sorting. When all positioning slots 501 are filled, it sends a signal to the main control system, and then the entire fixture is sent to the picking end 70 to wait for AGV or manual removal to proceed to the next process. The coil sorter 50 can receive coils that have been aligned and arrange them precisely and orderly in the positioning slots 501 to form a neat matrix. This greatly facilitates subsequent automatic welding, insertion, inspection, or packaging processes, improving the automation level of the entire production line. It collects and outputs data in "groups," and automatically sends it away once a tray is full, allowing for uninterrupted continuous production. It also facilitates docking with AGV carts or assembly lines, achieving seamless connection between processes.

[0053] In one embodiment, the working process of the high-speed inductor coil conditioning mechanism is as follows:

[0054] Step 1, Loading and Laying: The operator pours a large number of inductor coils into the vibratory feeder 201. The vibratory feeder 201 starts, and through its internal track and screening mechanism, it arranges the coils in an orderly manner and conveys them one by one to the material tray 202. The material tray 202 spreads the coils evenly through micro-vibration, laying them flat within the camera's field of view, ready to be "inspected".

[0055] Step 2, Visual Scanning and Recognition: The industrial camera is triggered to acquire an image of the entire material tray 202. The vision system software rapidly processes the image: first, it locates the circular outline of all coils and obtains their center coordinates; then, within each coil area, specific image algorithms, such as local binarization for fin features, accurately identify the orientation of the fins and calculate the angle that each coil needs to be rotated and corrected. The entire process is completed within milliseconds.

[0056] Step 3: Path Planning and Grasping: The vision system packages the coordinates and angles of all identified coils and sends them to the robotic arm control system. Based on algorithms such as the "nearest point principle," the control system plans the optimal path for the robotic arm to move and grasp, maximizing efficiency.

[0057] Step 4, Grasping and Aerial Correction: The robotic arm's drive module 401 moves the suction cup assembly 402 directly above the first target coil. The suction cup unit 4020 descends and picks up the coil. The instant the suction cup rises, its bottom rotating mechanism immediately activates, driving the suction cup and coil to rotate in a predetermined uniform direction based on angle data provided by the vision system. "Correcting while moving" or "correcting immediately after grasping" is the core of achieving high speed.

[0058] Step 5, Placement and Assembly: The robotic arm carries the calibrated coil to the target station above the coil arranger 50. The suction cup descends and precisely releases the coil into the positioning slot 501. Subsequently, the robotic arm immediately moves to the next target and repeats the gripping and calibration process.

[0059] Step Six: Loop and Output: Steps Two through Five are continuously looped until the robotic arm fills all the stations in the current row of the sorter when it picks up multiple coils at once. Then, the sorter automatically moves to one station to prepare for receiving the next row of coils. When all stations of the sorter are full, the system controls the sorter module to move it to the material handling end 70 and prompts for a material change. Simultaneously, a new empty fixture may be sent into the workstation. The entire process is seamlessly integrated, achieving high-speed, continuous, automated production.

[0060] See Figure 2 In one embodiment, the visual recognition system 30 has a built-in image processing module, which executes the following algorithm to identify the positive and negative directions and center coordinates of the inductor coil, including:

[0061] S10. Control the industrial camera to acquire images of the material tray 202, and use Gaussian filtering to denoise the images;

[0062] S20. Based on the denoised image, edge detection is performed using the Canny operator, and the circular outline of all inductor coils in the image is identified by the Hough circle transform algorithm. The center coordinates of the circular outline are calculated as the center coordinates of the inductor coil, and the first region image of each coil is located.

[0063] S30. Based on the first region image of each location, a local adaptive binarization method based on gray value variance is used to process it to generate a binarized image.

[0064] S40. Extract the largest connected component from the binarized image and label it as the fin region image; calculate the minimum bounding rectangle of the fin region image, and define the direction of the long side of the minimum bounding rectangle as the orientation of the inductor coil.

[0065] During image acquisition and preprocessing, the industrial camera, in conjunction with a specific light source such as a high-brightness ring-shaped white LED light source, triggers and acquires an image covering the entire working area of ​​the material tray 202. The acquired color image is usually immediately converted to a grayscale image to reduce the amount of data processing. Subsequently, Gaussian filtering is used to perform convolution processing on the grayscale image, effectively suppressing random noise in the image, such as camera sensor noise and ambient light interference.

[0066] To achieve coil localization and coarse center acquisition, the Canny operator is first used to extract edges from the preprocessed image. The Canny operator effectively preserves the circular edge gradient information of the coil. Then, the edge image is fed into the Hough circle transform algorithm. This algorithm detects contours in the image that conform to circular features and outputs the center coordinates and radius of each detected circle. Each detected circle represents an inductor coil, and its center coordinates are directly used as the center coordinates of that coil. Simultaneously, based on the center and radius of each circle, the algorithm crops a square region encompassing the entire coil from the original image. This region is the first region image, which is the region of interest for subsequent fine feature analysis.

[0067] Furthermore, to enhance and extract fin features, a local adaptive binarization algorithm is employed. Unlike global binarization, which uses a fixed threshold, this method calculates the threshold individually for each pixel in the image. The formula is typically: T = μ + k*σ, where T is the threshold, μ is the mean gray value within the pixel's neighborhood window, σ is the standard deviation, and k is a correction coefficient. Due to the height difference between the fins (flattened portion) and the circular coil body, a significant difference in brightness (e.g., the fins appear brighter or darker) will occur under specific lighting conditions. Local adaptive binarization can enhance this local contrast, dynamically calculating the optimal segmentation threshold, thereby robustly separating the fin region from the coil body and generating a clear black-and-white binarized image.

[0068] In the binarized image, the fins form a significant white or black connected region. By extracting the largest connected region, the location of the fins can be accurately pinpointed. The minimum bounding rectangle of this connected region is calculated. The direction of the long side of this rectangle naturally represents the extension direction of the fin, i.e., the flattened portion. The angle of this long side, such as the angle with the horizontal axis of the image, is defined as the orientation angle of the inductor coil, with 0° or 180° representing upright or reversed orientation. The entire image processing flow runs at high speed in an industrial control computer or embedded processor. The identified results are sent in real time to the control system of the robotic arm unit 40 via a communication interface, instructing it to complete the grasping and rotation correction actions.

[0069] This embodiment employs a strategy of "global localization followed by local fin feature recognition," avoiding interference and misjudgments caused by directly searching for minute features in the entire complex image, thus greatly improving the robustness and accuracy of the recognition algorithm. The use of local adaptive binarization effectively overcomes common industrial interference problems such as uneven illumination and coil body reflection, and can segment fin features more stably and accurately than the global thresholding method.

[0070] In one embodiment, the image processing module is further configured to perform the following steps to verify and correct the recognition result, including:

[0071] Pre-collect inductor coil image samples including upright, reverse, tilted, overlapping, and various lighting conditions to construct a training dataset; use the training dataset to train a convolutional neural network to obtain a classification verification model, the classification verification model is configured to take the cropped coil image as input and output the coil orientation classification result and confidence level;

[0072] The first region image of each coil is input into the classification verification model, and the coil orientation classification result and confidence score are output.

[0073] When the orientation of the output coil is consistent with the orientation determined according to the minimum bounding rectangle, the output result of the classification verification model is taken as the final orientation.

[0074] When the orientation of the output coil is inconsistent with the orientation determined by the minimum bounding rectangle, it is determined whether the confidence level is higher than the preset confidence threshold. If yes, the output result of the classification verification model is used as the final orientation. If no, the orientation determined by the minimum bounding rectangle is maintained as the final orientation, and the corresponding coil is marked as requiring manual re-inspection.

[0075] First, a large number of inductor coil images from real production environments were systematically collected. These images included not only normal forward and reverse playback samples, but also various abnormal situations, including:

[0076] Tilt: The coil was not fully vibrated.

[0077] Overlap: The two coils partially overlap.

[0078] Abnormal lighting: overexposure, underexposure, localized shadows, reflections.

[0079] Foreign object interference: There is oil, metal shavings, etc. on the tray.

[0080] Defective products: poorly flattened fins, deformed coils.

[0081] Then, the orientation (positive / negative) of the coils in all these images is precisely labeled manually, forming a high-quality labeled dataset. Using this dataset, a lightweight convolutional neural network, such as MobileNetV2, is trained. The learning objective of this model is: to take a cropped image of a coil as input and output its orientation classification (positive / negative) and a confidence score representing the degree of prediction certainty.

[0082] After the model is trained, the verification and decision-making steps are executed. While the traditional algorithm processes the image, the "first region image" of each cropped coil is simultaneously fed into the deployed CNN classification and verification model. The system then obtains two results: the traditional algorithm result, the orientation determined by the minimum bounding rectangle of the fin; and the deep learning result, the orientation predicted by the CNN model and its confidence level. When the output coil orientation matches the orientation determined by the minimum bounding rectangle, this consistent result is adopted as the final orientation. This demonstrates that the two algorithms mutually verify each other, and the results are reliable. When the orientation of the output coil is inconsistent with the orientation determined by the minimum bounding rectangle, the relationship between the confidence level and the preset confidence threshold is further evaluated. If the confidence level is higher than the threshold, it indicates that the CNN model has given a different judgment from the traditional algorithm with a very high degree of confidence. This usually means that the traditional algorithm misjudged due to some kind of interference, while the CNN made the correct decision due to its generalization ability. Therefore, the output result of the classification and verification model is taken as the final orientation. Conversely, if the confidence level is lower than the threshold, the orientation determined by the minimum bounding rectangle is maintained as the final orientation, and the corresponding coil is marked as requiring manual re-inspection. That is, the system cannot make a reliable judgment on its own and human intervention is required.

[0083] Traditional image processing algorithms rely on predefined rules, such as finding circles or the largest connected component. They are prone to failure when encountering extreme anomalies not considered during training, such as severe reflections or singular overlaps. Deep learning models, by learning from massive amounts of anomalous samples, possess powerful generalization capabilities to handle these unexpected situations, effectively correcting the misjudgments of traditional algorithms. Furthermore, the system no longer relies on a single algorithm for decision-making. The two algorithms cross-validate each other, significantly reducing the overall misjudgment rate and ensuring the accuracy of the system's output.

[0084] Figure 3 A schematic diagram of the high-speed inductor coil straightening mechanism according to another embodiment is provided. For example... Figure 3 As shown, in one embodiment, there are two vibratory feeding units 20, which are symmetrically arranged on the left and right sides of the frame 10, and the coil arranger 50 is arranged between the two vibratory feeding units 20.

[0085] Two vibratory feeding units 20 can simultaneously and independently screen, sort, and convey coils. This effectively doubles the feeding speed, providing twice the material source for subsequent visual recognition and grasping, fundamentally solving the problem that a single feeding unit might become a bottleneck for the entire production line efficiency. When there are two vibratory feeding units 20, there are also two corresponding visual recognition systems (industrial cameras), such as... Figure 3 As shown.

[0086] The robotic arm, typically a gantry structure, sits atop two feed trays 202 on the left and right, and a central sorter. This layout minimizes the average distance and directness of the robotic arm's movement from one feed tray 202 to the central sorter. Compared to a linear layout where materials are fed from one end of the equipment and sorted from the other, this symmetrical layout significantly reduces the robotic arm's ineffective movement time, further improving the overall operating cycle time.

[0087] Preferably, the suction cup assembly 402 includes four independently controlled suction cup units 4020, forming a quadruple suction cup assembly 402, such as... Figure 1 As shown. In the quadruple suction cup assembly 402, the rotation angle of each suction cup unit 4020 is independently controlled by the coil orientation identified by the vision recognition system 30.

[0088] In this embodiment, the four-suction cup assembly 402 is fixedly connected to the Z-axis slider of the robotic arm gantry or drive module 401 via a mounting plate. The assembly comprises four completely independent suction cup units 4020, typically evenly arranged on the mounting plate. Each unit includes:

[0089] Vacuum suction cup: Typically made of flexible silicone or polyurethane material, it is used to generate negative pressure to pick up the coil. Its size must match the size of the coil.

[0090] Small servo motors: the core component enabling independent rotation. Each suction cup is directly mounted on the output shaft of a servo motor, driven by the motor for precise rotational movement.

[0091] Mini cylinder or electric actuator: Drives the entire suction cup unit 4020 (including servo motor and suction cup) to perform independent vertical telescopic movements to achieve gripping and release.

[0092] Air tubing and electrical wiring: Each suction cup is connected to an independent vacuum line and cables that power the servo motor and cylinder and control signals.

[0093] The robot's control system needs to be capable of controlling multiple auxiliary axes. Each servo motor and cylinder acts as an independent control axis. The system is equipped with multiple independent vacuum generators or a single vacuum generator with multiple solenoid valves to achieve independent on / off control of the vacuum in each suction cup. Using the quadruple suction cup assembly 402, the robot can complete the gripping, alignment, and placement of four coils in a single stroke, theoretically achieving nearly four times the efficiency of a single suction cup system. This directly achieves a "high-speed" sorting effect. The process is completed synchronously during lifting and movement after gripping, without occupying any additional cycle time. Compared to the traditional "align first, then grip" or "grip and place on a alignment table for secondary positioning" approach, the process is extremely simplified. Furthermore, since the four suction cups can extend and retract independently, even if the four coils are at slightly different heights on the tray 202, simultaneous and reliable gripping can be ensured. For example, assuming a sparse area of ​​coils, the vision system may only identify two or three qualified coils. The control system can flexibly activate only two or three suction cups to work, without waiting for all four to be gathered, further improving equipment utilization.

[0094] In one embodiment, a piezoelectric ceramic vibrator is installed at the bottom of the material tray 202, and the piezoelectric ceramic vibrator is connected to a controller;

[0095] The controller is configured to first drive the vibrator at a first frequency to concentrate the inductor coils, and then drive the vibrator at a second frequency higher than the first frequency to spread the inductor coils out.

[0096] In this embodiment, the first frequency is relatively low, and the vibrator operates in a concentrated mode. The lower frequency vibration primarily generates larger amplitudes and longer energy waves. This vibration mode makes it easier for the material to move as a whole and generate centripetal force. Within the limited space of the material tray 202, the coils will gradually gather towards the center or low-lying areas of the tray 202, as if being "shaked" together, thus solving the problem of reduced grasping efficiency due to an initially sparse distribution. The second frequency is higher than the first frequency, and the vibrator operates in a spread-out mode. The higher frequency vibration primarily generates high-acceleration, low-amplitude micro-movements. This high-frequency micro-vibration effectively overcomes the static friction and adhesion between particles, allowing each coil to obtain independent energy, thus "jumping" and dispersing on the surface of the tray 202. Simultaneously, high-frequency vibration helps the coils find a stable equilibrium position using their own shape (circular), ultimately achieving uniform spreading, avoiding stacking and obstruction, and creating optimal conditions for visual recognition. The piezoelectric ceramic vibrator is typically attached or installed at the bottom of the tray 202. Piezoelectric ceramic elements have the characteristics of fast response speed, high control precision and low noise. When a high-frequency electrical signal is applied, it will produce precise micro deformation, thereby driving the material tray 202 to vibrate.

[0097] The "spread-out mode" ensures that each coil is independently laid out in the field of view, greatly reducing misidentification or recognition failure caused by coils contacting or obstructing each other, which is a prerequisite for achieving high-precision recognition. The "centralized mode" avoids the coils being sparsely distributed on the edge of the material tray 202, ensuring that the coils are all within the reachable and efficient grasping range of the robot, reducing the long-distance movement of the robot due to grasping targets that are too far away, and shortening the overall cycle time.

[0098] Figure 4 A partially enlarged structural schematic diagram of the coil arranger 50 is provided. (See attached diagram.) Figure 4 As shown, in one embodiment, the coil arranger 50 is provided with a plurality of positioning slots 501 that match the gripping layout of the robotic arm unit 40. The positioning slots 501 include coil center positioning pins 5010 and coil fin limiters 5011.

[0099] Specifically, the positioning groove 501 base is typically a recessed cavity, its diameter and depth slightly larger than the diameter and thickness of the coil body, used to accommodate and support the circular main body of the coil, providing support in the Z direction. The coil center positioning pin 5010 is vertically fixed at the center of the positioning groove 501 base, its diameter fitting with the circular hole at the center of the coil with a small clearance; for example, if the diameter of the circular hole is 1.0 mm, the diameter of the positioning pin is 0.95 mm, allowing for easy insertion while ensuring accurate center positioning. The coil fin limiter 5011 can be implemented as follows... Figure 1 The preferred type is the trough type.

[0100] Implementation Form 1 (Slot Type): A limiting slot matching the shape, thickness, and length of the coil fins is milled into the edge of the positioning slot 501 substrate. When the coil is inserted, its fins must be able to be fully embedded in this slot.

[0101] Implementation Method Two (Pin Type): Two small pins are installed at the edge of the positioning groove 501 substrate, located on both sides of the fin. The gap between these two pins is exactly equal to the thickness of the fin, thus "clamping" the fin in the middle and restricting its position.

[0102] Implementation Form 3 (Block Type): Install a movable L-shaped limiting block. After the coil is inserted, push the block with a cylinder or manually to press down the fins, which serves to limit and fix them.

[0103] During coil arrangement, the robotic arm moves the coil, which has already undergone aerial alignment and is now oriented uniformly, to above the arranger. The robotic arm descends along its Z-axis, guiding the coil toward the positioning slot 501. The center positioning pin first inserts into the circular hole at the center of the coil, completing the initial alignment. As it continues to descend, the coil's fins slide into or embed into the fin limiters. The coil body falls completely into the positioning slot 501, at which point the coil is fully constrained in all six degrees of freedom (X, Y, Z, θ), and its position is precisely fixed. The suction cup breaks the vacuum, releasing the coil, and the robotic arm lifts up, ready for the next placement.

[0104] The shape of the positioning slot 501 perfectly matches the physical contour of the inductor coil. The number and layout of the positioning slots 501 can be precisely matched to the number and layout of the robotic suction cups. The sorter can be designed to receive 4, 8 or more coils at a time. Like a mold, the coil can only be placed in the slot in a single correct orientation, thus ensuring that the orientation and position height of all coils are consistent. Primary positioning (coarse positioning) is achieved by the coil center positioning pin 5010. It uses the circular hole in the center of the coil to first roughly fix the coil in the correct position (XY coordinates), completing the center alignment. Final positioning (fine positioning) is achieved by the coil fin limiter 5011. It holds the coil's unique fin structure, firstly ensuring that the circumferential rotation angle of the coil is fixed (i.e., uniform orientation); secondly, as a mistake-proofing mechanism, if the coil orientation is incorrect (e.g., fins facing down instead of upward), the fins will not be able to be placed in the limiting slot, and the coil will not be fully positioned or will be stuck, thus immediately exposing the problem.

[0105] Therefore, by using mechanical hard-positioning, the absolute position and orientation of each coil in the coil arranger are completely consistent. The built-in error-proofing mechanism ensures that if the robotic arm fails to rotate and correct, or if it grabs a coil in the wrong orientation for any reason, the coil will not be correctly placed (it may be tilted, stuck, or even fall off) because the fins cannot fall into the limit switch. This will be immediately detected by the equipment's sensors, triggering an alarm and stopping the machine, preventing batch quality accidents. During the movement or transfer of the coil arranger to the next station, the firmly fixed coils will not shift or collide due to vibration or inertia, ensuring quality stability from arrangement to the next process.

[0106] In one embodiment, the discharge track of the vibratory feeder 201 is provided with a guide 60, such as... Figure 3 As shown. The width of the guide 60 is greater than the diameter of a single inductor coil but less than the sum of the diameters of two inductor coils, so as to control the inductor coils to be output in a single, orderly manner.

[0107] The discharge track of the vibratory feeder is a narrow channel. The guide 60, through its narrow width, forces the passing inductors to pass, allowing only those that meet specific conditions to pass, using the physical size (diameter D) of the inductors as the screening criterion. Let the width of the guide 60 be W, its design follows a simple mathematical inequality: D < W < 2D; when W > D, it ensures that an inductor can pass smoothly regardless of its orientation (standing or lying down), as long as its maximum projected size is less than W. When W < 2D, it ensures that two inductors side-by-side can never pass simultaneously. If two coils attempt to pass in parallel, their total width will exceed W, causing them to get stuck at the entrance of the guide 60 until vibration separates them. This mechanical structure achieves the goal of allowing the coils to pass "one by one" in a queue.

[0108] The vibratory feeder operates, and the coils are vibrated and conveyed to the discharge port along the internal track. Upon entering the discharge track, the coils may be individual, overlapping, or side-by-side. When the coils reach the guide 60, physical sorting begins: individual coils pass smoothly through the narrow channel. Two overlapping coils, with the upper coil potentially being scraped back into the hopper due to instability. Two side-by-side coils are firmly blocked outside the channel because their total width 2D > W. Subsequent vibrations continuously attempt to "knead" these two coils into a front-to-back alignment until one enters the channel first, followed closely by the other. After passing the guide 60, all coils are forcibly arranged into a stable single-file column and orderly conveyed to the subsequent tray 202.

[0109] Understandably, without the guide 60, the parallel coils might get stuck in narrower sections of the track or at bends, causing the entire line to stop and requiring manual intervention. The guide 60 eliminates these potential problems at the very beginning, ensuring continuous and smooth production and reducing unplanned downtime. The single-row, orderly output means a stable and predictable feeding rate, facilitating matching with the downstream robotic arm's cycle time.

[0110] Preferably, the frame 10 is provided with a material picking end 70 at the end of the coil arranger 50, such as... Figure 3 As shown. Once the positioning slot 501 on the coil arranger 50 is filled, the arranged inductor coil is sent to the picking end 70, waiting to be transported to the next process.

[0111] After the robotic arm places the last coil into the sorter positioning slot 501, the full-material detection sensor is triggered or the counter determines that the quantity is full. Upon receiving the full-material signal, the PLC controls the robotic arm to stop feeding material into the sorter. The PLC then issues a command to start the transfer mechanism, transporting the fully loaded sorter fixture to the picking end 70. Once the positioning sensor at the picking end 70 detects that the fixture has been precisely positioned and locked, it sends a feedback signal to the PLC. The PLC then notifies external equipment such as AGVs, robots, or operators via an audible and visual alarm, the MES system, or I / O signals that "material is ready and can be picked up." After the external equipment removes the fully loaded fixture, the transfer mechanism returns the empty fixture to its working position, ready to begin the next cycle. This method allows the sorting equipment to begin sorting the next fixture immediately after completing one without stopping and waiting. The picking operation is completed in parallel by another device, greatly reducing equipment idle time and improving equipment utilization.

[0112] The above description covers all specific embodiments provided by this invention. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented using electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention.

Claims

1. A high-speed coil straightening mechanism, characterized in that, include: frame; At least one vibratory feeding unit is disposed on the frame; each vibratory feeding unit includes a vibratory feeder and a material tray; the material tray is disposed at the outlet of the vibratory feeder and is configured to lay out the inductor coil by vibration; A visual recognition system includes an industrial camera positioned directly above the tray, the industrial camera being configured to capture images of inductor coils laid flat on the tray; The robotic arm unit includes a translational drive module mounted on the frame and a suction cup assembly driven by the drive module; the suction cup assembly includes at least one independently controlled suction cup unit, each of which can independently perform vertical telescopic movement and rotational movement around its own axis. A coil arranger, mounted on the frame, is used to receive and arrange inductor coils with uniform orientation; The visual recognition system is communicatively connected to the robotic arm unit, used to identify the orientation and coordinates of each inductor coil, and to control the robotic arm unit to grasp and correct the inductor coil before placing it in the coil arranger; The visual recognition system has a built-in image processing module, which executes the following algorithm to identify the positive and negative directions and center coordinates of the inductor coil, including: The industrial camera is controlled to capture images of the material tray, and Gaussian filtering is used to denoise the images; Based on the denoised image, edge detection is performed using the Canny operator, and the circular outlines of all inductor coils in the image are identified using the Hough circle transform algorithm. The center coordinates of the circular outlines are calculated as the center coordinates of the inductor coils, and the first region image of each coil is located. Based on the first region image of each location, a local adaptive binarization method based on gray value variance is used to process it and generate a binarized image. Extract the largest connected component from the binarized image and label it as the fin region image; calculate the minimum bounding rectangle of the fin region image, and define the direction of the long side of the minimum bounding rectangle as the orientation of the inductor coil; The image processing module is further configured to perform the following steps to verify and correct the recognition results, including: Pre-collect inductor coil image samples including upright, reverse, tilted, overlapping, and various lighting conditions to construct a training dataset; use the training dataset to train a convolutional neural network to obtain a classification verification model, the classification verification model is configured to take the cropped coil image as input and output the coil orientation classification result and confidence level; The first region image of each coil is input into the classification verification model, and the coil orientation classification result and confidence score are output. When the orientation of the output coil is consistent with the orientation determined according to the minimum bounding rectangle, the output result of the classification verification model is taken as the final orientation. When the orientation of the output coil is inconsistent with the orientation determined by the minimum bounding rectangle, it is determined whether the confidence level is higher than the preset confidence threshold. If yes, the output result of the classification verification model is used as the final orientation. If no, the orientation determined by the minimum bounding rectangle is maintained as the final orientation, and the corresponding coil is marked as requiring manual re-inspection.

2. The high-speed inductor coil straightening mechanism according to claim 1, characterized in that, The number of vibratory feeding units is two, symmetrically arranged on the left and right sides of the frame, and the coil arranger is arranged between the two vibratory feeding units.

3. The high-speed inductor coil straightening mechanism according to claim 1, characterized in that, The suction cup assembly includes four independently controlled suction cup units, forming a four-joint suction cup assembly.

4. The high-speed inductor coil straightening mechanism according to claim 3, characterized in that, The rotation angle of each suction cup unit in the four-joint suction cup assembly is independently controlled by the coil orientation identified by the visual recognition system.

5. The high-speed inductor coil straightening mechanism according to claim 1, characterized in that, A piezoelectric ceramic vibrator is installed at the bottom of the material tray, and the piezoelectric ceramic vibrator is connected to the controller; The controller is configured to first drive the vibrator at a first frequency to concentrate the inductor coils, and then drive the vibrator at a second frequency higher than the first frequency to spread the inductor coils out.

6. The high-speed inductor coil straightening mechanism according to claim 1, characterized in that, The coil arranger is provided with several positioning slots that match the gripping layout of the robotic arm unit. The positioning slots include coil center positioning pins and coil fin limiters.

7. The high-speed inductor coil straightening mechanism according to claim 1, characterized in that, The vibratory feeder has a guide on its discharge track. The width of the guide is greater than the diameter of a single inductor coil but less than the sum of the diameters of two inductor coils, so as to control the inductor coils to be output in a single, orderly manner.

8. The high-speed inductor coil straightening mechanism according to claim 1, characterized in that, The frame is provided with a material picking end at the end of the coil arranger. When the positioning slot on the coil arranger is filled, the arranged inductor coil is sent to the material picking end and waits to be transported to the next process.

Images