Intelligent control method and system for crucible winding

Abstract

This application relates to the field of crucible preform preparation technology, specifically to an intelligent control method and system for crucible piercing, applied to a winding mechanism and a piercing mechanism controlled by a host computer, wherein a switching platform is deployed between the winding mechanism and the piercing mechanism, including the following steps: obtaining the basic parameters of the current crucible preform, determining the working parameters of the winding mechanism and the piercing parameters of the piercing mechanism; determining a first control command based on the working parameters, and determining a second control command based on the first control command and the piercing parameters; when the winding mechanism executes the first control command and the piercing mechanism executes the second control command to process the crucible preform, acquiring a processing image of the crucible preform, and performing defect image recognition; when at least one defect feature exists, identifying at least one erroneous operation that matches the defect feature, optimizing the first control command based on the erroneous operation, and simultaneously optimizing the second control command while optimizing the first control command.

Description

Technical Field

[0001] This invention relates to the field of crucible preform preparation technology, and in particular to an intelligent control method and system for crucible burr wrapping. Background Technology

[0002] A crucible preform is a pre-fabricated crucible, typically made of refractory material, used for melting metals at high temperatures or for other chemical reactions. The advantage of crucible preforms is their convenience and speed; multiple crucibles can be prepared in advance for unforeseen needs. In use, simply place the preform into the furnace and heat it.

[0003] Currently, in the fabrication of crucible preforms, fibers or other reinforcing materials are first wound onto a mandrel, arranging the reinforcing materials at specific angles and densities to obtain a high-strength, high-temperature-resistant preform. Then, needles are used to puncture and fix the reinforcing materials, further increasing their strength and density. During the crucible preform fabrication process, the needle-punching process allows the reinforcing materials to bond more tightly, improving the overall strength and temperature resistance of the crucible preform. Simultaneously, the needle-punching process can also improve the interlayer bonding force of the reinforcing materials, reducing interlayer delamination and further enhancing the reliability and stability of the crucible preform. However, existing crucible preform winding and puncturing equipment has a low degree of automation and long operation time.

[0004] The invention with patent number CN202010546987.2 proposes a carbon fiber crucible needle-punching forming equipment. The technical solution adopted is as follows: the equipment includes a bottom needle-punching device, a side needle-punching device, a shaping support device, a movable support device, and an atomizing spray device. The bottom needle-punching device corresponds to the bottom position of the crucible preform and performs needle-punching forming treatment on the bottom of the crucible preform. The side needle-punching device corresponds to the side position of the crucible preform and performs needle-punching forming treatment on the side of the crucible preform. The shaping support device corresponds to the bottom needle-punching device and the side needle-punching device and supports the crucible preform. The side needle-punching device and the shaping support device are respectively located on the movable support device. The movable support device drives the side needle-punching device and the shaping support device to adjust their positions. The atomizing spray device corresponds to the shaping support device and sprays liquid onto the crucible preform placed on the shaping support device. The aforementioned device is highly automated, but it does not inspect the quality of the products when they are fed into the oven for curing, which will affect the stability and continuity of the production line.

[0005] Therefore, the present invention provides an intelligent control method and system for crucible burr wrapping. Summary of the Invention

[0006] This invention provides an intelligent control method and system for crucible winding and piercing. Based on set parameters, working parameters and piercing parameters are generated to control the winding mechanism and piercing mechanism to perform winding and piercing operations on the crucible model to obtain a crucible preform. After the piercing operation is completed, the quality of the current crucible preform is detected, which can detect defects, inhomogeneities and other problems. This method can not only improve product quality, but also improve production efficiency.

[0007] In a first aspect, an intelligent control method for crucible piercing is provided, applied to a winding mechanism and a piercing mechanism controlled by a host computer, wherein a switching platform is deployed between the winding mechanism and the piercing mechanism, comprising the following steps:

[0008] Obtain the basic parameters of the current crucible preform, and determine the working parameters of the winding mechanism and the needle-punching parameters of the needle-punching mechanism;

[0009] Based on the working parameters, a first control command is determined, and based on the first control command and the needle-punching parameters, a second control command is determined; wherein, the first control command is used to control the winding mechanism to perform a winding operation, and the second control command is used to control the needle-punching mechanism to perform a needle-punching operation coupled with the winding operation;

[0010] When the winding mechanism executes the first control command and the needle punching mechanism executes the second control command to process the crucible preform, the processing image of the crucible preform is acquired, and defect image recognition is performed.

[0011] When at least one defect feature exists, at least one erroneous operation that conforms to the defect feature is determined, and the first control instruction is optimized based on the erroneous operation. When optimizing the first control instruction, the second control instruction is optimized simultaneously.

[0012] When no defect features are present, the control robot arm moves the processed crucible preform.

[0013] In conjunction with the first aspect, obtaining the basic parameters of the current crucible preform and determining the working parameters of the winding mechanism and the acupuncture parameters of the acupuncture mechanism includes:

[0014] Obtain the basic parameters of the crucible preform and perform process modeling; among which...

[0015] Process modeling includes: dimensional modeling, winding modeling, and needle punch density modeling;

[0016] Based on dimensional modeling and winding modeling, the number of winding turns, the rotational speed and tension of the winding mechanism are calculated to obtain the working parameters of the winding mechanism;

[0017] Based on size modeling and needle density modeling, the depth and spacing of needles are calculated to obtain the needle parameters of the needle-punching mechanism.

[0018] In conjunction with the first aspect, after the winding mechanism executes the first control command, it further includes:

[0019] When a feedback signal indicating completion of winding is received from the winding mechanism, the control parameters for switching platforms are determined.

[0020] Based on control parameters, a rotation command is generated to control the motor of the switching platform, thereby rotating the crucible preform to a position below the needle-punching mechanism; wherein,

[0021] Feedback signals and rotation commands are triggered in association, and the control parameters in the rotation commands are determined by direct mapping.

[0022] In conjunction with the first aspect, the defect image recognition includes:

[0023] Preprocess the current image to be detected;

[0024] Key features are extracted from the preprocessed current image to be detected using histogram of oriented gradients.

[0025] The extracted key features are input into the defect recognition model for identification to determine whether the surface of the crucible preform is uniform.

[0026] When the surface of the crucible preform is uneven, determine the location of the defect.

[0027] In conjunction with the first aspect, after the defect image recognition, it further includes:

[0028] Obtain the recognition results of defect image identification, and determine the defect features and erroneous operations;

[0029] Based on the defect characteristics and erroneous operations, a compensation plan is determined; the compensation plan includes repair compensation and optimization compensation.

[0030] The repair compensation mechanism determines the compensation scheme based on the mesh modeling mechanism and the differential compensation mechanism of the defect location, and is triggered when the first control command and the second control command corresponding to the defect location have consecutive errors: where,

[0031] The mesh modeling mechanism is used to determine the abnormal working parameters and abnormal needle-punching parameters corresponding to each mesh at the defect location;

[0032] The differential compensation mechanism is used to determine the compensation coefficients for abnormal working parameters and abnormal acupuncture parameters;

[0033] The optimization compensation is based on a linear tuning mechanism and a linear iteration mechanism according to the defect location, and is triggered when an operational error occurs in the first or second control command corresponding to the defect location; wherein,

[0034] The linearization optimization mechanism is used to build a linear constraint curve based on the working parameters and needle punching parameters of the crucible preform. The linear constraint curve is continuously iterated and calculated through a linear iteration mechanism to determine the linear constraint curve. The linear constraint curve is then used as the target constraint curve for crucible bodies of the same specifications for constraint compensation.

[0035] In conjunction with the first aspect, during the winding and needle-punching process, the acquisition of processing images of the crucible preform also includes:

[0036] Based on the processing images, the position and angle information of the crucible preform are acquired in real time, and the needling and winding processes are monitored in real time; among them...

[0037] The real-time monitoring results are displayed through a scrolling window, which scrolls synchronously with the position and angle information of the crucible preform in the same time series.

[0038] Adjust the winding operation parameters of the first control command or the needle puncture operation parameters of the second control command based on the monitoring results.

[0039] In conjunction with the first aspect, the synchronous scrolling also includes:

[0040] The acquired position / angle information is compared with a preset threshold. When the position / angle information does not match the preset threshold, an alarm message is generated and displayed on the host computer.

[0041] In conjunction with the first aspect, the host computer display includes:

[0042] Determine the required chart type based on the type of location and angle information;

[0043] Adjust the elements displayed in the chart based on the distribution and characteristics of location and angle information;

[0044] Map the location and angle information onto the coordinate axes of the chart type to obtain the target chart;

[0045] The target chart is rendered and displayed on the host computer.

[0046] In conjunction with the first aspect, the defect image recognition also includes:

[0047] When at least two defect features exist, calculate the correlation coefficient between the two defect features;

[0048] When the correlation coefficient is greater than the preset correlation value, it indicates that the two defect features are from the same source and executes the first optimization instruction. The first optimization instruction is used to stop the first control instruction for the winding mechanism and the second control instruction for the needle punching mechanism, and extract the key parameters of the first control instruction and the second control instruction to perform gradient descent optimization.

[0049] When the correlation coefficient is less than the preset correlation value, it indicates that the two defects are independent erroneous operations, and the second optimization instruction is executed. The second optimization instruction is used to cross-calibrate the optimized first control instruction and the second control instruction after synchronous optimization.

[0050] Secondly, this application proposes an intelligent control system for crucible winding and acupuncture, applied to a winding mechanism and acupuncture mechanism controlled by a host computer, wherein a switching platform is deployed between the winding mechanism and the acupuncture mechanism, and the system includes:

[0051] Parameter setting module: used to obtain the basic parameters of the current crucible preform, and to determine the working parameters of the winding mechanism and the needle punching parameters of the needle punching mechanism;

[0052] Instruction setting module: used to determine the first control instruction based on the working parameters, and to determine the second control instruction based on the first control instruction and the needle punching parameters; wherein, the first control instruction is used to control the winding mechanism to perform the winding operation, and the second control instruction is used to control the needle punching mechanism to perform the needle punching operation coupled with the winding operation;

[0053] Defect recognition module: When the winding mechanism executes the first control command and the needle punching mechanism executes the second control command to process the crucible preform, the module acquires processing images of the crucible preform and performs defect image recognition; wherein,

[0054] When at least one defect feature exists, at least one erroneous operation that conforms to the defect feature is determined, and the first control instruction is optimized based on the erroneous operation. When optimizing the first control instruction, the second control instruction is optimized simultaneously.

[0055] When no defect features are present, the control robot arm moves the processed crucible preform.

[0056] The beneficial effects of this application are as follows:

[0057] This application proposes a collaborative control method during crucible preform processing, using winding parameters as a baseline and dynamically adapting needle-punching parameters to achieve coordinated control and resolve process defects and processing failures that may be caused by independently setting parameters. Defect identification is performed after image acquisition, and the winding and needle-punching parameters under coordinated control are synchronously corrected in a closed loop to prevent the limitations of offline defect detection. Simultaneously, because the coordinated control of parameter settings and defect detection feed back to the system control parameter settings, closed-loop unmanned control is achieved, improving processing efficiency.

[0058] This application achieves automated control of the crucible winding process, reduces manual intervention, and improves production efficiency and the stability of crucible preform quality.

[0059] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings.

[0060] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0061] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0062] Figure 1 This is a flowchart of the intelligent control method for crucible burr wrapping in an embodiment of the present invention;

[0063] Figure 2 This is a topology diagram of device control in an embodiment of the present invention;

[0064] Figure 3 This is a system composition diagram of an intelligent control system for crucible entanglement according to an embodiment of the present invention.

[0065] Figure label:

[0066] 10 is the host computer, 20 is the winding mechanism, 30 is the needle-punching mechanism, and 40 is the switching platform. Detailed Implementation

[0067] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0068] The fabrication of crucible preforms is a composite material molding process; see [link / reference]. Figure 2 The process is mainly carried out through a winding mechanism 20 and a needle-punching mechanism 30. Both mechanisms are controlled by a host computer 10, and the process is connected via a switching platform 40. The crucible preform winding mechanism 20 typically drives a servo motor and a yarn guiding system via a main shaft that rotates the mold. The yarn guiding system consists of a yarn guide nozzle, a tension controller, and a yarn frame. The core component of the needle-punching mechanism 30 is the needle plate. The needle plate is controlled by a servo drive system to reciprocate up and down, and the needle depth is adjusted by a needle depth adjustment module that adjusts the eccentric wheel or ball screw.

[0069] In the processing of the crucible preform, the crucible mold is first mounted on the spindle, and then the coaxiality of the mold center with the spindle is calibrated using a laser alignment instrument. During processing, the spindle rotates at a low speed, and the yarn guide rises spirally from the bottom of the crucible mold, winding and tightly fitting against the mold. The movement speed of the yarn guide is matched with the spindle speed. After winding is complete, the needle punching mechanism 30 punctures the wound crucible mold at a certain frequency through needle plates, driving fiber interweaving to generate the crucible preform. However, in the actual processing, needle breakage defects caused by independent parameter settings are prone to occur. The crucible preform can only be inspected after processing, and quality control cannot directly optimize the processing.

[0070] During the processing of the crucible preform, the winding tension, spindle speed, and winding angle of the winding mechanism 20 are set according to historical processes. Similarly, the needle density, needle depth, and needle plate feed speed of the needle punching mechanism 30 are set according to historical processes. Setting these two parameters independently leads to the following problem during processing: the winding tension causes excessively high carbon fiber layer compaction density; and the needle depth, needle plate feed speed, and needle density, when inserted into the excessively dense fiber layer, cause needle bending and breakage. In the presence of broken needles, because the needle plate contains numerous dense hooks, it will not be detected without manual inspection, potentially resulting in an entire batch of defective crucible preforms.

[0071] During the inspection of the processed crucible preform, the encoder of the feed motor of the needle punching mechanism 30 drifted, resulting in insufficient needle punching depth or insufficient tension in the winding mechanism 20, leading to loose winding. This resulted in irregular fiber layer extensions beyond the crucible edge area. For this type of quality defect, it can only be determined whether the crucible preform has processing quality defects after production is complete. Therefore, there is a problem of delayed detection, and defects may even exist in batches.

[0072] In the process of processing crucible preforms, if errors occur, the common methods are to reduce the winding tension (to solve the problem of layered bubbles) and increase the needle punching density (to solve the problem of insufficient density), but both methods will have some defects to some extent.

[0073] To address the aforementioned issues, this application proposes an intelligent control method for crucible piercing. During the crucible preform processing, the piercing parameters are dynamically adapted to the piercing parameters based on the piercing parameters, achieving coordinated control and resolving process defects and processing failures that may be caused by independently setting parameters. Defect identification is performed after image acquisition, and the piercing and piercing parameters under coordinated control are synchronously corrected in a closed loop to prevent the limitations of offline defect detection. Simultaneously, because the coordinated control of parameter settings and defect detection feed back to the parameter settings of the system control, closed-loop unmanned control is achieved, improving processing efficiency.

[0074] The solutions in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0075] Example 1:

[0076] See Figure 1 , Figure 1 A flowchart of a method for intelligent control of crucible burr wrapping.

[0077] Step S100: Obtain the basic parameters of the current crucible preform, and determine the working parameters of the winding mechanism 20 and the needle-punching parameters of the needle-punching mechanism 30;

[0078] The basic parameters are the physical properties of the crucible preform, such as its dimensions and materials, as well as its strength and density. By pre-modeling the winding mechanism 20 and the needle-punching mechanism 30, the initial process parameters for winding and needle-punching are automatically calculated, eliminating the need for manual or independent settings. However, the parameters generated by the process modeling may deviate from their intended purpose during the actual production of the winding mechanism 20 and the needle-punching mechanism 30. Therefore, only the initial process parameters can be determined. Since the crucible preform is processed by winding first and then needle-punching, the needle-punching parameters are determined based on the performance of the semi-finished crucible preform achieved by the winding mechanism 20.

[0079] In one embodiment, parameters are calculated by setting the diameter, height, carbon fiber type, and target density of the crucible. First, the actual required winding tension and spindle speed during winding are determined by using the tension range, spindle speed range, fiber modulus, and fiber diameter of the winding mechanism 20, and the fiber guiding speed is matched. Then, the needle punching parameters, including the needle punching density and depth, are calculated through coupling. For example, if there are ten layers of fiber during winding, the needle punching depth needs to penetrate all ten layers of fiber.

[0080] Step S200: Determine the first control command based on the working parameters, and determine the second control command based on the first control command and the needle-punching parameters; wherein, the first control command is used to control the winding mechanism 20 to perform the winding operation, and the second control command is used to control the needle-punching mechanism 30 to perform the needle-punching operation coupled with the winding operation;

[0081] When sending control commands to the winding mechanism 20 and the needle punching mechanism 30, a dynamic coupling mechanism between the winding master command and the needle punching slave command is established. The winding mechanism 20 and the needle punching mechanism 30 are linked in real time through the industrial bus, so that the winding state and the needle punching operation are matched.

[0082] In one embodiment, the winding mechanism 20 continuously winds the fiber according to the set tension, rotation speed, and cross-winding angle based on the operating parameters. When a certain layer is wound, if the real-time thickness of the fiber layer is higher than the calculated theoretical value, it is determined that this may be due to fiber elastic deformation or a deviation in the operating parameters of the winding mechanism 20. At this point, if the preparation requirements of the crucible preform are met, the winding result of the crucible preform obtained through the first instruction, combined with the needle-punching parameters, is used to change the needle-punching depth. The needle-punching trajectory is corrected so that the needle-punching trajectory and depth can be coupled with the winding angle and thickness. Through a cross-grid method on the crucible preform, the final crucible performance requirements are achieved.

[0083] Step S300: When the winding mechanism 20 executes the first control command and the needle punching mechanism 30 executes the second control command to process the crucible preform, the processing image of the crucible preform is acquired and defect image recognition is performed;

[0084] During the processing, images of the crucible preform at each stage are collected to determine whether there are process defects or actual product defects of the crucible preform, such as blistering, broken needles, or insufficient needle density. These images are then compared with preset thresholds to determine the specific defect characteristics.

[0085] In one embodiment, the presence of bubbling, broken needles, and actual product defects, i.e., morphological defects on the crucible preform, is determined by capturing images at a frequency of 0.01 seconds per image. Furthermore, by calculating the needle density and number of needle strikes in the continuously captured images, it is determined whether there is insufficient needle density, thus identifying individual defects.

[0086] In one embodiment, based on the captured image, a preset feature recognition model is used to identify that a certain fiber layer has blistering and that there is insufficient needle-punching density in some areas, which causes blistering, i.e., there is a composite defect.

[0087] Step S400: When at least one defect feature exists, at least one erroneous operation that conforms to the defect feature is determined, and the first control instruction is optimized according to the erroneous operation, and the second control instruction is optimized simultaneously while optimizing the first control instruction;

[0088] In one embodiment, based on the identification result of the defect image, if it is a single defect, the cause of the single defect, i.e., a single erroneous operation, is determined to be either the winding mechanism 20 or the needle-punching mechanism 30. If it is an erroneous operation of the needle-punching mechanism 30, the second control command is corrected, and compensation processing can be performed if necessary to repair it. If it is an error of the winding mechanism 20, the erroneous operation of the first control command is corrected, and the second control command is optimized simultaneously.

[0089] In one embodiment, when multiple defects exist, i.e. composite defects, each defect corresponds to a separate erroneous operation. If a defect exists that is a defect of the winding mechanism 20, the first control command of the winding mechanism 20 is optimized first, and then the second control command of the needle-punching mechanism 30 is optimized simultaneously. Then, it is determined again whether the second control command after the first optimization can correct the defect caused by the needle-punching mechanism 30. If it still cannot correct the defect caused by the needle-punching mechanism 30, the second control command is optimized again.

[0090] In one embodiment, if a characteristic defect exists, but this defect is caused by multiple erroneous operations, an operation chain is formed according to the operational sequence of these erroneous operations. When the erroneous operation is caused by the winding mechanism 20, according to the operation chain, the control parameters of the first erroneous operation of the first control command of the winding mechanism 20 are first optimized. Then, each erroneous operation is corrected sequentially according to the operation chain. After each erroneous operation is corrected, the second control command needs to be synchronously optimized, and it is determined whether there are still process errors. If not, the current operating parameters are used as the target control parameters. If so, the correction is continuously performed according to the operation chain.

[0091] Step S500: When there are no defect features, control the unloading robot arm to move the processed crucible preform.

[0092] After defect image recognition, if the prepared crucible preform is free of defects, a transfer command containing the crucible position coordinates will be sent to the feeding robot arm.

[0093] Example 2:

[0094] To provide precise process parameters for the first and second control commands based on the basic parameters of the crucible preform, the following operations are performed in conjunction with the winding mechanism 20 and the needle punching mechanism 30 before the crucible preform is prepared:

[0095] Obtain the basic parameters of the crucible preform and perform process modeling; among which...

[0096] Process modeling includes: dimensional modeling, winding modeling, and needle punch density modeling;

[0097] The winding mechanism 20 and the needle punching mechanism 30 of this application are connected to a host computer 10. The host computer 10 calculates the working parameters of the winding mechanism 20 and the needle punching parameters of the needle punching mechanism 30. After receiving the basic parameters, the host computer 10 performs multi-dimensional process modeling based on the process flow of the winding mechanism 20 and the needle punching mechanism 30, automatically calculates the crucible parameters and operating parameters, quantifies the process requirements of the crucible preform, and the equipment control parameters.

[0098] In one embodiment, the host computer 10 automatically calculates the working parameters of the winding machine, such as tension and rotation speed, by having the operator input the geometric parameters, material parameters, and performance parameters of the crucible preform on the HMI page. Based on the quality (expected quality) of the crucible preform semi-finished product output by the winding mechanism 20, it determines the acupuncture parameters of the acupuncture mechanism 30, such as acupuncture density and acupuncture depth. In this process, no manual parameters are required, and the time to obtain the working parameters and acupuncture parameters is based on the computing power of the host computer 10, which greatly improves work efficiency and avoids errors caused by manual settings.

[0099] Based on the dimensional modeling and winding modeling, the number of winding turns, the rotational speed and tension of the winding mechanism 20 are calculated to obtain the working parameters of the winding mechanism 20;

[0100] During the calculation of the working parameters of the winding mechanism 20 by the host computer 10, three-dimensional modeling is achieved based on the three-dimensional dimensions of the crucible preform, the winding path, and the needle punching density, using a CAD model. A three-dimensional mesh model is then generated to calculate the coverage area and number of layers of the winding. Next, winding modeling is performed, and a mechanical model of the winding tension and fiber deformation is established by combining the elastic modulus and elongation at break of the fiber material. This determines the spindle speed and tension during the winding process.

[0101] In one embodiment, by using 3D modeling and winding calculations, the process accuracy is improved, which can prevent insufficient tension and dimensional errors caused by manually setting parameters.

[0102] Based on size modeling and needle density modeling, the depth and spacing of the needles are calculated to obtain the needle parameters of the needle mechanism 30.

[0103] During the calculation of the needle punching parameters of the needle punching mechanism 30 by the host computer 10, based on the number of fiber layers and deformation of the crucible preform semi-finished product achieved by the working parameters of the winding mechanism 20, and combined with the target density and fiber volume fraction, a mapping relationship between the number of needle punches and the degree of fiber interlacing is constructed to determine the needle punching parameters.

[0104] In one embodiment, the fiber interlacing requirements are combined with size modeling and wall thickness layering data (thickness per layer) and needled density modeling, and the needled depth and needled density are calculated by spatial geometry algorithms, resulting in higher needled accuracy.

[0105] Example 3:

[0106] Regarding the preparation of the crucible preform, when the winding mechanism 20 and the needle-punching mechanism 30 are connected in the process, because they are different mechanisms, there is generally a detection interval when they operate independently with a single parameter setting. In this application, defect detection runs throughout the entire process, so no detection interval is needed. Apart from the detection interval, there will inevitably be process delays and potential accuracy errors when connecting processes.

[0107] When a feedback signal indicating completion of winding is received from the winding mechanism 20, the control parameters of the switching platform 40 are determined.

[0108] The needle-punching mechanism 30 automatically switches the platform 40 to the needle-punching mechanism 30 when the winding operation of the winding mechanism 20 is completed, based on the comparison table of feedback signals and control parameters, without the need for manual intervention.

[0109] In one embodiment, unlike the traditional winding mechanism 20 and needle punching mechanism 30 which are both PLC controlled, there is no delay or process stoppage during the process connection, which improves the overall preparation efficiency.

[0110] A rotation command is generated based on control parameters to control the motor of the switching platform 40, thereby rotating the crucible preform to a position below the needle-punching mechanism 30; wherein,

[0111] Feedback signals and rotation commands are triggered in association, and the control parameters in the rotation commands are determined by direct mapping.

[0112] The switching platform 40 adopts a signal edge triggering and parameter encoding method. When the switching platform 40 receives the feedback signal from the winding mechanism 20, the rising edge directly triggers the rotation command. The rotation angle and speed in the rotation command are adaptively executed through a preset mapping table. The associated triggering method allows the rotation command to obtain control parameters without calculation, thus achieving zero-delay process connection.

[0113] In one embodiment, the rising edge of the feedback signal of the winding mechanism 20 triggers the rotation command of the switching platform 40. The rotation angle and speed determined by the processing image, as well as the feedback signal of the preset mapping table and the triggering of the rotation angle and speed, enable the motor encoder to be directly mapped according to the control parameters and rotated below the needle punching mechanism 30 without any delay. The needle punching misalignment problem can also be solved by the processing image.

[0114] Example 4:

[0115] In order to improve the efficiency of defect detection in the processing image of crucible preforms, subtle features of the crucible preforms are identified during the defect image recognition process to improve the accuracy of defect identification.

[0116] Preprocess the current image to be detected;

[0117] Preprocessing utilizes multimodal image enhancement to eliminate interference from mechanical vibrations and fiber texture noise that occur during the fabrication of crucible preforms inside the plant. Dynamic contrast stretching improves the identification of defect details and maintains smooth edges to prevent blurred edge defects.

[0118] In one embodiment, modal image enhancement is used to improve image clarity, making it easier to extract defect features.

[0119] Key features are extracted from the preprocessed current image to be detected using histogram of oriented gradients.

[0120] The directional gradient histogram reflects the boundary orientation of the defect in the gradient direction, and the gradient magnitude is used to quantify the grayscale difference between the defect area and the normal area. By using block-level histograms, key features free from local noise are generated, enabling the identification of minute defects.

[0121] In one embodiment, when the current image to be detected is processed by horizontal and vertical histograms, the gradient magnitude of the defect region is higher than that of the normal fiber texture. By constructing the directional histogram and achieving block-level normalization, the angle of fiber winding is corresponding to the HOG histogram of the normal region on the image, while the multi-directional gradient of the defect region represents the bubble edge, thus achieving the capture of minute features.

[0122] The extracted key features are input into the defect recognition model for identification to determine whether the surface of the crucible preform is uniform.

[0123] In the surface uniformity detection of crucible preforms, after the key features are input into the defect recognition model, the underlying edge texture information displayed on different feature images of the crucible preforms is used to determine whether the crucible preforms are uniform by the uniformity of the edge texture.

[0124] When the surface of the crucible preform is uneven, determine the location of the defect.

[0125] The defect location is the uneven area on the surface of the crucible preform, which also includes errors in the process flow.

[0126] In one embodiment, when the surface of the crucible preform is uneven, the defect location is precisely located using a thermal map and edge frame based on the identification results, thereby improving the accuracy of defect feature identification.

[0127] Example 5:

[0128] During the processing of crucible preforms, defects were identified through defect image recognition. A scheme for precise positioning and targeted compensation was proposed to address how to compensate for and optimize these defects.

[0129] Obtain the recognition results of defect image identification, and determine the defect features and erroneous operations;

[0130] Based on the identification results of the defect image, the host computer 10 correlates the visual shape, size, and grayscale distribution of the crucible preform in the defect image with the winding tension and needle penetration depth of the winding mechanism 20 and the needle penetration mechanism 30 in the manufacturing process. This identifies the erroneous operation that caused the corresponding defect. Through the mapping between defects and operations, the error can be directly deduced from the defect, enabling both the detection of operational errors and the tracing of their origins.

[0131] In one embodiment, for defect image recognition, during the needle-punching operation of the crucible preform, there is a ring-shaped bubble cluster. The bubble generation is mainly due to uneven force and insufficient needle-punching depth. Based on the corresponding processing sequence data of the winding mechanism 20 and the needle-punching mechanism 30, it can be determined that during the winding of the winding mechanism 20, there is a winding tension fluctuation, and during the needle-punching operation, there is insufficient needle-punching depth, resulting in the final ring-shaped bubble cluster.

[0132] Based on the defect characteristics and erroneous operations, a compensation plan is determined; the compensation plan includes repair compensation and optimization compensation.

[0133] Repair and compensation are for continuous operational errors, such as sensor drift and deviations caused by mechanical wear, which require the elimination of defects by correcting local parameters.

[0134] The optimization compensation is designed to address occasional operational errors, such as instantaneous voltage fluctuations or insufficient needle penetration depth in a particular instance. It prevents the recurrence of similar errors through iterative optimization of global parameters.

[0135] In one embodiment, if the erroneous operations are related, different compensation schemes are executed for different erroneous operations.

[0136] The repair compensation mechanism determines the compensation scheme based on the mesh modeling mechanism and the differential compensation mechanism of the defect location, and is triggered when the first control command and the second control command corresponding to the defect location have consecutive errors: where,

[0137] The mesh modeling mechanism is used to determine the abnormal working parameters and abnormal needle-punching parameters corresponding to each mesh at the defect location;

[0138] The differential compensation mechanism is used to determine the compensation coefficients for abnormal working parameters and abnormal acupuncture parameters;

[0139] In one embodiment, the mesh modeling mechanism divides the surface of the crucible preform into multiple mesh units, locates the mesh using defect coordinates, and determines whether there are continuous errors using the parameters of the winding mechanism 20 of the first control command and the needle-punching mechanism 30 of the second control command.

[0140] In one embodiment, after identifying the erroneous operation and determining the degree of deviation from the error, a compensation coefficient is calculated based on the degree of deviation of the erroneous operation, with a greater compensation force for a larger deviation.

[0141] In one embodiment, spatial gridding is used to bind the defect location and control command parameters, perform multi-grid collaborative compensation, and perform differentiated compensation based on the degree of deviation, so that there will be no undercompensation or overcompensation during compensation.

[0142] The optimization compensation is based on a linear tuning mechanism and a linear iteration mechanism according to the defect location, and is triggered when an operational error occurs in the first or second control command corresponding to the defect location; wherein,

[0143] The linearization optimization mechanism is used to build a linear constraint curve based on the working parameters and needle punching parameters of the crucible preform. The linear constraint curve is continuously iterated and calculated through a linear iteration mechanism to determine the linear constraint curve. The linear constraint curve is then used as the target constraint curve for crucible bodies of the same specifications for constraint compensation.

[0144] The linearization optimization mechanism uses data on nonlinear process parameters such as winding tension, winding speed, number of winding layers, and needle penetration depth to fit (least square method) into a linear model. Through incremental iteration of the linear iteration mechanism, the model prediction value and the target value are brought close together, generating a stable linear constraint curve. During the preparation of the crucible preform, the production parameters are continuously optimized.

[0145] In one embodiment, the needle penetration depth in the second control command is affected by a single deviation caused by instantaneous voltage fluctuations, triggering optimization compensation. A linear model of tension and needle penetration depth is established to determine density uniformity. Based on the linear model, if it is determined that density uniformity needs to be improved, the tension coefficient, depth coefficient, and constant term of subsequent operations will be adjusted. This not only optimizes the current crucible preform but also optimizes the subsequent production of new crucible preforms.

[0146] Example 6:

[0147] In order to improve the accuracy of defect image recognition during the acquisition of processing images, dynamic tracking is performed by continuously rolling the crucible preform and combining it with the spatiotemporal display of the scrolling window. This allows for the discovery of defects, their display, and compensation optimization.

[0148] Based on the processing images, the position and angle information of the crucible preform are acquired in real time, and the needling and winding processes are monitored in real time; among them...

[0149] The real-time monitoring results are displayed through a scrolling window, which scrolls synchronously with the position and angle information of the crucible preform in the same time series.

[0150] The detection of processed images mainly relies on binocular vision positioning. Triangulation is used to calculate feature points on the crucible surface to determine the real-time position of the crucible preform. A contour matching algorithm is used to determine the center of the crucible end face (to address the issue of edge deviation that can occur with large crucibles). A rotary encoder is used to calculate the real-time rotation angle, and the position and angle data are compared with a preset trajectory. An alarm is triggered when the deviation exceeds a threshold.

[0151] In one embodiment, frame images of the crucible surface are acquired using binocular vision devices installed on both sides of the switching platform 40. A feature point recognition algorithm is then used to identify N stable feature points. The coordinates of the entire crucible preform are calculated based on these stable feature points to determine the position and angle information during processing. On the host computer 10, the processing image of the crucible preform is displayed synchronously in real-time via a scrolling window. This synchronous scrolling tracks the dynamic image of the crucible preform's processing. Because synchronous scrolling eliminates blind spots during monitoring, defects are detected in real-time, preventing defects from appearing on edges or the surface of the crucible preform.

[0152] In one embodiment, a scrolling window is displayed on the host computer 10's display interface, containing a real-time image area and a parameter curve area, and scrolls and refreshes with the processing timeline. The two areas are aligned by timestamps. Image frames, position coordinates, and angles are aligned at the same point in time. During comparison, if a deviation is found, clicking on the deviation point of the curve will backtrack to display the image at the corresponding moment and show the corresponding erroneous operation.

[0153] Adjust the winding operation parameters of the first control command or the needle puncture operation parameters of the second control command based on the monitoring results.

[0154] In one embodiment, based on the monitoring results, it is determined whether the parameters of the winding mechanism 20 or the needle-punching mechanism 30 need to be adjusted, and then the consequences of the corresponding erroneous operation are eliminated by modifying the first control command or the second control command.

[0155] Example 7:

[0156] In the process of acquiring processing images, in order to accurately display the difference between the actual processing trajectory of the crucible preform and the preset processing curve on the host computer 10, it is necessary to determine whether there is any erroneous operation.

[0157] The acquired position / angle information is compared with a preset threshold. If the position / angle information does not match the preset threshold, an alarm message is generated and displayed on the host computer 10.

[0158] The preset thresholds include static and dynamic thresholds set according to the processing requirements. During the comparison process on the host computer 10, a dual-channel parallel approach is used to achieve millisecond-level real-time comparison at the hardware layer and deviation analysis at the software layer to trigger alarms when thresholds are exceeded. Based on the deviation, corresponding alarm information is generated and pushed to the host computer 10 interface in real time via the OPCUA protocol, making processing anomalies preventable, controllable, and traceable, reducing the scrap rate caused by position / angle deviations.

[0159] Example 8:

[0160] During the acquisition of processing images, the visualization interface of the host computer 10 employs the following methods to achieve statistical display of errors in batches of crucible preforms or individual chart analysis and recording of errors in a single crucible preform:

[0161] Determine the required chart type based on the type of location and angle information;

[0162] Based on positional and angular information, and according to characteristics, the axial deviation of the crucible preform over time, the angular deviation of the crucible preform with processing height, and the standard positional deviation of batch-manufactured crucible preforms can be determined. By adapting corresponding charts, the spatial distribution of angular deviation is displayed through a 3D scatter plot, the axial deviation over time is displayed through a line trend chart, and the positional deviation of batch manufacturing is displayed through a heat map, thus achieving automatic display of different types of information.

[0163] In one embodiment, as shown in the table below, automatic chart adaptation prevents information display discrepancies:

[0164]

[0165] Adjust the elements displayed in the chart based on the distribution and characteristics of location and angle information;

[0166] In one embodiment, the elements displayed by the chart include coordinate axes, legends, colors, and markers. Distribution analysis can identify data clusters and outlier detection can mark outliers. Element adjustments include scaling the coordinate axis range, and deviations are displayed using color coding to show different degrees of deviation. Outliers can be magnified to show their corresponding defects. This prevents traditional chart displays from being overloaded with information and causing key information to be overlooked due to the large amount of content on the chart.

[0167] Map the location and angle information onto the coordinate axes of the chart type to obtain the target chart;

[0168] In this embodiment, the location information includes the three-dimensional coordinates of the crucible preform's position and the angle information during processing. The three-dimensional coordinates are mapped to a Cartesian coordinate system, and the angle information is mapped to a polar coordinate system. By combining the same coordinate axis of the two coordinate systems, heterogeneous angle and position data are displayed simultaneously, indicating that an alarm may exist in the corresponding area.

[0169] The target chart is rendered and displayed on the host computer 10.

[0170] In this embodiment, the target icon is rendered and displayed on the display interface of the host computer 10 through hardware accelerated rendering and multi-screen collaborative display technology, thereby achieving synchronous display and rendering and improving the display effect.

[0171] Example 9:

[0172] In the process of defect image recognition, when a defect is discovered, in order to make a more accurate judgment, avoid blind adjustments, and avoid coupling interference, the following possible implementation method exists in defect image recognition:

[0173] When at least two defect features exist, calculate the correlation coefficient between the two defect features;

[0174] A defect feature coupling algorithm is used to quantify the correlation between various defects. Feature parameters such as defect area, perimeter, grayscale mean, and edge gradient are extracted from defect images to form a feature vector matrix. Then, a pre-selected similarity algorithm is used to calculate the correlation coefficient between two defect features. By calculating the correlation coefficient using both feature vectors and spatial distance, this algorithm addresses situations where loose fiber layers and excessively deep needle penetration result in spatially adjacent defects that might be judged as originating from the same source.

[0175] When the correlation coefficient is greater than the preset correlation value, it indicates that the two defect features are from the same source and executes the first optimization instruction; wherein, the first optimization instruction is used to stop the first control instruction of the winding mechanism 20 and the second control instruction of the needle punching mechanism 30, and extract the key parameters of the first control instruction and the second control instruction, and perform gradient descent optimization.

[0176] Same-source operation is based on the mapping rules between defect type and equipment action to determine the main responsible control instruction that caused the defect;

[0177] In one embodiment, the winding deviation of the winding mechanism 20 is mapped by the looseness of the fiber layer. The insufficient needle penetration depth of the needle penetration mechanism 30 is mapped by the needle penetration depth deviation.

[0178] Key parameter extraction employs a random forest feature algorithm, selecting at least three parameters with the highest contribution from multiple parameters in the control command. A defect feature parameter is used as the loss function, and loss function convergence control is performed.

[0179] In one embodiment, by stopping the main instruction, the key parameters are optimized in a targeted manner, preventing the problem of coupling interference between the two instructions when they exist simultaneously.

[0180] When the correlation coefficient is less than the preset correlation value, it indicates that the two defects are independent erroneous operations, and the second optimization instruction is executed. The second optimization instruction is used to cross-calibrate the optimized first control instruction and the second control instruction after synchronous optimization.

[0181] The second optimization instruction solves the parameter silo problem by decoupling and coordinating the two instructions through independent optimization and cross-calibration.

[0182] In one embodiment, particle swarm optimization (PSO) is performed on the entanglement / needling control commands respectively, with the objective function being the characteristic parameters of the corresponding defects, for example: entanglement defects → area minimization, needled defects → depth deviation minimization;

[0183] In one embodiment, based on an LSTM coupling error prediction model, the LSTM coupling error prediction model takes the optimized parameter changes as input and outputs a coupling error compensation value. When the parameter adjustment of the two commands is greater than ±5%, the coupling term is dynamically compensated, for example, the needle penetration depth drift caused by the increase of winding tension. This prevents the first control command from increasing to decrease tension and the second control command from increasing to adjust the needle penetration depth, which would cause coupling interference between the two commands and result in parameter defects even after synchronization.

[0184] See Figure 3 This application also provides an intelligent control system for crucible entanglement, applied to a winding mechanism 20 and a needle-punching mechanism 30 controlled by a host computer 10. A switching platform 40 is deployed between the winding mechanism 20 and the needle-punching mechanism 30. The system includes:

[0185] Parameter setting module: used to obtain the basic parameters of the current crucible preform, and to determine the working parameters of the winding mechanism 20 and the needle punching parameters of the needle punching mechanism 30;

[0186] Instruction setting module: used to determine the first control instruction based on the working parameters, and to determine the second control instruction based on the first control instruction and the needle punching parameters; wherein, the first control instruction is used to control the winding mechanism 20 to perform the winding operation, and the second control instruction is used to control the needle punching mechanism 30 to perform the needle punching operation coupled with the winding operation;

[0187] Defect recognition module: When the winding mechanism 20 executes the first control command and the needle punching mechanism 30 executes the second control command to process the crucible preform, the module acquires processing images of the crucible preform and performs defect image recognition; wherein,

[0188] When at least one defect feature exists, at least one erroneous operation that conforms to the defect feature is determined, and the first control instruction is optimized based on the erroneous operation. When optimizing the first control instruction, the second control instruction is optimized simultaneously.

[0189] When no defect features are present, the control robot arm moves the processed crucible preform.

[0190] It is understood that all relevant content of each step involved in the above method embodiments can be referenced in the embodiments of the intelligent control system for crucible entanglement, and will not be repeated here.

[0191] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for intelligent control of crucible winding and acupuncture, applied to a winding mechanism and acupuncture mechanism controlled by a host computer, wherein a switching platform is deployed between the winding mechanism and the acupuncture mechanism, characterized in that, Includes the following steps: Obtain the basic parameters of the current crucible preform, and determine the working parameters of the winding mechanism and the needle-punching parameters of the needle-punching mechanism; Based on the working parameters, a first control command is determined, and based on the first control command and the needle-punching parameters, a second control command is determined; wherein, the first control command is used to control the winding mechanism to perform a winding operation, and the second control command is used to control the needle-punching mechanism to perform a needle-punching operation coupled with the winding operation; When the winding mechanism executes the first control command and the needle punching mechanism executes the second control command to process the crucible preform, the processing image of the crucible preform is acquired, and defect image recognition is performed. When at least one defect feature exists, at least one erroneous operation that conforms to the defect feature is identified, and the first control instruction is optimized based on the erroneous operation. When optimizing the first control instruction, the second control instruction is optimized simultaneously. When no defect features are present, the control robot arm moves the processed crucible preform.

2. The intelligent control method for crucible burr wrapping as described in claim 1, characterized in that, The process of obtaining the basic parameters of the current crucible preform and determining the working parameters of the winding mechanism and the needle-punching parameters of the needle-punching mechanism includes: Obtain the basic parameters of the crucible preform and perform process modeling; among which... Process modeling includes: dimensional modeling, winding modeling, and needle punch density modeling; Based on dimensional modeling and winding modeling, the number of winding turns, the rotational speed and tension of the winding mechanism are calculated to obtain the working parameters of the winding mechanism; Based on size modeling and needle density modeling, the depth and spacing of needles are calculated to obtain the needle parameters of the needle-punching mechanism.

3. The intelligent control method for crucible burr wrapping as described in claim 1, characterized in that, After the winding mechanism executes the first control command, it further includes: When a feedback signal indicating completion of winding is received from the winding mechanism, the control parameters for switching platforms are determined. Based on control parameters, a rotation command is generated to control the motor of the switching platform, thereby rotating the crucible preform to a position below the needle-punching mechanism; wherein, Feedback signals and rotation commands are triggered in association, and the control parameters in the rotation commands are determined by direct mapping.

4. The intelligent control method for crucible burr wrapping as described in claim 1, characterized in that, The defect image recognition includes: Preprocess the current image to be detected; Key features are extracted from the preprocessed current image to be detected using histogram of oriented gradients. The extracted key features are input into the defect recognition model for identification to determine whether the surface of the crucible preform is uniform. When the surface of the crucible preform is uneven, determine the location of the defect.

5. The intelligent control method for crucible burr wrapping as described in claim 1, characterized in that, After the defect image is identified, the method further includes: Obtain the recognition results of defect image identification, and determine the defect features and erroneous operations; Based on the defect characteristics and erroneous operations, a compensation plan is determined; the compensation plan includes repair compensation and optimization compensation. The repair compensation mechanism determines the compensation scheme based on the mesh modeling mechanism and the differential compensation mechanism of the defect location, and is triggered when the first control command and the second control command corresponding to the defect location have consecutive errors: where, The mesh modeling mechanism is used to determine the abnormal working parameters and abnormal needle-punching parameters corresponding to each mesh at the defect location; The differential compensation mechanism is used to determine the compensation coefficients for abnormal working parameters and abnormal acupuncture parameters; The optimization compensation is based on a linear tuning mechanism and a linear iteration mechanism according to the defect location, and is triggered when an operational error occurs in the first or second control command corresponding to the defect location; wherein, The linearization optimization mechanism is used to build a linear constraint curve based on the working parameters and needle punching parameters of the crucible preform. The linear constraint curve is continuously iterated through a linear iteration mechanism to determine the target constraint curve of the crucible body of the same specification. The crucible preform is then constrained and compensated using the target constraint curve.

6. The intelligent control method for crucible burr wrapping as described in claim 1, characterized in that, During the winding and needle-punching process, the acquisition of processing images of the crucible preform also includes: Based on the processing images, the position and angle information of the crucible preform are acquired in real time, and the needling and winding processes are monitored in real time; among them... The real-time monitoring results are displayed through a scrolling window, which scrolls synchronously with the position and angle information of the crucible preform in the same time series. Adjust the winding operation parameters of the first control command or the needle puncture operation parameters of the second control command based on the real-time monitoring results.

7. The intelligent control method for crucible burr wrapping as described in claim 6, characterized in that, The synchronous scrolling also includes: The acquired position / angle information is compared with a preset threshold. When the position / angle information does not match the preset threshold, an alarm message is generated and displayed on the host computer.

8. The intelligent control method for crucible burr wrapping as described in claim 7, characterized in that, The host computer display includes: Determine the required chart type based on the type of location and angle information; Adjust the elements displayed in the chart based on the distribution and characteristics of location and angle information; Map the location and angle information onto the coordinate axes of the chart type to obtain the target chart; The target chart is rendered and displayed on the host computer.

9. The intelligent control method for crucible burr wrapping as described in claim 1, characterized in that, The defect image recognition also includes: When at least two defect features exist, calculate the correlation coefficient between the two defect features; When the correlation coefficient is greater than the preset correlation value, it indicates that the two defect features are from the same source and executes the first optimization instruction; wherein, the first optimization instruction is used to stop the first control instruction for the winding mechanism and the second control instruction for the needle punching mechanism, and extract the key parameters of the first control instruction and the second control instruction, and perform gradient descent optimization. When the correlation coefficient is less than the preset correlation value, it indicates that the two defects are independent erroneous operations, and the second optimization instruction is executed. The second optimization instruction is used to cross-calibrate the optimized first control instruction and the second control instruction after synchronous optimization.

10. An intelligent control system for crucible entanglement and acupuncture, applied to a winding mechanism and acupuncture mechanism controlled by a host computer, wherein a switching platform is deployed between the winding mechanism and the acupuncture mechanism, characterized in that, The system includes: Parameter setting module: used to obtain the basic parameters of the current crucible preform, and to determine the working parameters of the winding mechanism and the needle punching parameters of the needle punching mechanism; Instruction setting module: used to determine the first control instruction based on the working parameters, and to determine the second control instruction based on the first control instruction and the needle punching parameters; wherein, the first control instruction is used to control the winding mechanism to perform the winding operation, and the second control instruction is used to control the needle punching mechanism to perform the needle punching operation coupled with the winding operation; Defect recognition module: When the winding mechanism executes the first control command and the needle punching mechanism executes the second control command to process the crucible preform, the module acquires processing images of the crucible preform and performs defect image recognition; wherein, When at least one defect feature exists, at least one erroneous operation that conforms to the defect feature is identified, and the first control instruction is optimized based on the erroneous operation. When optimizing the first control instruction, the second control instruction is optimized simultaneously. When no defect features are present, the control robot arm moves the processed crucible preform.

Images