A duct lip casting control method and system based on infrared positioning

Abstract

The application discloses a method and system for controlling a tunnel lip casting based on infrared positioning, and relates to the technical field of lip casting control. The method for controlling the tunnel lip casting based on infrared positioning projects infrared reference light spots into the inner cavity of the tunnel lip and collects infrared images; through background correction and sub-pixel extraction, actual contour point sets and light spot projection centers are obtained, the center is taken as a reference to generate a positioning eccentricity vector by constraint fitting and to perform positioning adjustment, after the adjustment, infrared time sequence signals of each orientation sprue are collected by sliding, a filling infrared radiation change rate curve is generated by radiation and flow mapping, an asymmetric compensation reference value is generated according to the eccentricity vector, the spatial relative deviation degree is analyzed, the metal liquid pressure injection flow value is generated by composite mapping processing, and the tunnel lip casting control processing is performed. The application realizes the precise control of the lip wall thickness and improves the forming quality of the casting by performing the tunnel lip casting control processing based on the metal liquid pressure injection flow value.

Description

Technical Field

[0001] This invention relates to the field of duct casting control technology, specifically to a duct lip casting control method and system based on infrared positioning. Background Technology

[0002] Infrared positioning technology is a non-contact measurement technology that uses infrared light as an information carrier to determine the spatial position or geometric features of a target object through projection, reception, and image analysis. In the field of precision casting, the spatial positioning accuracy of the mold directly affects the dimensional consistency and wall thickness uniformity of the casting. For thin-walled castings with complex internal cavity structures, such as ducted fan housings, the wall thickness of the lip part is usually only on the order of millimeters, and the requirements for the concentricity of the mold cavity are strict. During the casting process, the placement of the mold, mold closing, and subsequent pouring all involve the determination and adjustment of spatial orientation. Therefore, effective positioning methods are needed to ensure the relative positional accuracy between the components.

[0003] The limitations of the existing technology include at least the following problems: When using plaster molds for integrated casting of duct lips, the anisotropic shrinkage behavior of the plaster mold during the drying stage, and the random distribution of shrinkage of different individuals, results in an uncertain eccentric offset between the actual center of the inner cavity and the theoretical center. Before the outer frame is closed, this eccentricity is difficult to be quantitatively perceived and specifically corrected. After the mold is closed, the position of the plaster mold is rigidly locked, and the eccentricity cannot be mechanically corrected in subsequent processes. This causes the concentricity deviation between the lip mold cavity and the main body mold cavity to be directly converted into uneven distribution of the circumferential wall thickness of the casting. At the same time, the secondary expansion of the plaster mold caused by heat during the high-temperature pouring stage is affected by the internal density difference formed by the initial shrinkage, exhibiting asymmetric thermal deformation related to the individual shrinkage characteristics. This exacerbates the dynamic offset of the mold cavity position, making it difficult to stably control the wall thickness deviation within a stable range throughout the entire process by relying solely on static positioning before pouring. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method and system for controlling the casting of duct lip based on infrared positioning, which solves the problems of difficulty in correcting eccentricity and stable control of lip wall thickness deviation caused by gypsum mold shrinkage and thermal deformation.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a duct lip casting control method based on infrared positioning, comprising the following steps: projecting an infrared reference spot with a preset geometric shape into the inner cavity of the duct lip, and acquiring an infrared band image containing the infrared reference spot and the duct lip; based on the preset infrared radiation background characteristics of the gypsum material, performing background grayscale non-uniformity correction on the infrared band image, and extracting the actual contour point set of the duct lip and the projected geometric center of the infrared reference spot at sub-pixel level; using the projected geometric center as a fitting reference, performing constraint fitting processing on the actual contour point set to generate a positioning eccentricity vector, and further... Positioning adjustment is performed; after positioning adjustment, molten metal is poured and filled into the duct lip, and infrared radiation time-series signals corresponding to the gates in different directions of the duct lip are collected by sliding acquisition. After thermal radiation and flow rate correlation mapping processing, the filling infrared radiation change rate curves of the gates in each direction are generated; based on the positioning eccentricity vector, the anisotropic shrinkage characteristics are determined, and the asymmetric compensation reference value of the gates in each direction is generated, and the spatial relative deviation between the filling infrared radiation change rate curves is analyzed; based on the spatial relative deviation and the asymmetric compensation reference value, after composite mapping processing, the molten metal hydraulic injection flow rate value of the gates in each direction is generated, and the casting control processing of the duct lip is performed.

[0006] Furthermore, the specific steps for background gray-level non-uniformity correction of infrared band images are as follows: acquire a set of gypsum background calibration images and perform filtering processing; construct a surface model based on the filtered gypsum background calibration image set and store it as a preset infrared radiation background feature of gypsum material; calculate the background estimate pixel by pixel of the infrared band image based on the surface model, and perform background compensation on the infrared band image based on the background estimate to complete the background gray-level non-uniformity correction.

[0007] Further, the specific steps for extracting the actual contour point set of the inner edge of the duct lip and locating the projected geometric center of the infrared reference spot are as follows: For the infrared band image after background gray-level non-uniform correction, the sub-pixel edge detection operator based on Zernike moments is used to extract the edge, and the candidate edge point set of the inner edge of the duct lip is obtained; connected component analysis is performed on the candidate edge point set, and morphological closing operation is performed based on the connected component analysis results to obtain the actual contour point set of the inner edge of the duct lip; adaptive threshold segmentation is performed on the region where the infrared reference spot is located in the infrared band image to extract the initial region of the spot, and the first moment and zero moment of the initial region of the spot are calculated using the gray-level centroid method, and the projected geometric center of the infrared reference spot is located.

[0008] Further, the specific steps for generating the positioning eccentric vector are as follows: taking the projection geometric center as the initial fitting circle center, the actual contour point set of the inner cavity is subjected to coarse error elimination to obtain the filtered contour point set; based on the filtered contour point set and combined with the preset constraint conditions, a constrained least squares circle fitting objective function is established, and the optimal circle center coordinates are generated by solving the problem; the optimal circle center coordinates and the projection geometric center are subjected to vector subtraction to generate the positioning eccentric vector.

[0009] Further, the specific steps for generating the filling infrared radiation rate of change curves for gates in each direction are as follows: First-order difference operations are performed on the infrared radiation time-series signal to obtain discrete sequences of infrared radiation intensity change rates for gates in each direction; these discrete sequences are input into a preset radiation-flow correlation mapping model for radiation-flow transformation processing, outputting a sequence of filling rate characterization values ​​for gates in each direction; spline interpolation is performed on the filling rate characterization value sequence to generate the filling infrared radiation rate of change curves for gates in each direction.

[0010] Furthermore, the preset steps of the radiation-flow correlation mapping model are as follows: Collect time-series signals of the gate infrared radiation from multiple sets of duct lip samples under different metal hydraulic jet flow rates, and simultaneously record the actual metal hydraulic jet flow rate curves to form a radiation-flow calibration dataset; divide the radiation-flow calibration dataset into windows and calculate the infrared radiation intensity change rate samples and flow rate sample labels; using the infrared radiation intensity change rate samples as independent variables and the flow rate sample labels as dependent variables, establish the radiation-flow correlation mapping model using a data-driven nonlinear regression method; and use cross-validation and parameter optimization strategies to determine the optimal structural parameters of the radiation-flow correlation mapping model.

[0011] Further, the specific steps for generating the asymmetric compensation reference values ​​for gates in each direction are as follows: The positioning eccentricity vector is decomposed to extract anisotropic shrinkage characteristics, including the eccentricity direction angle and eccentricity distance. The direction of the gate with the maximum resistance corresponding to the eccentricity direction angle is determined, and the gate with the maximum resistance direction is taken as the reference compensation direction. The compensation weight coefficient for each gate is calculated using the reference compensation direction as the center and in combination with the eccentricity distance. The compensation weight coefficient is then corrected with the preset base metal hydraulic injection flow rate setting value to obtain the asymmetric compensation reference values ​​for each gate.

[0012] Further, the specific steps for analyzing the spatial relative deviation between the filling infrared radiation rate of change curves of gates in different directions are as follows: Integrate the filling infrared radiation rate of change curves over time to construct a spatial filling distribution vector; using the gate direction pointed to by the eccentricity angle as the reference direction, sort the spatial filling distribution vectors according to the gate direction to form a sequence of directions and cumulative filling amounts; extract the cumulative filling amount characterization value in the reference direction and the cumulative filling amount characterization value in the direction perpendicular to the reference direction from the sequence of directions and cumulative filling amounts; generate the spatial relative deviation based on the cumulative filling amount characterization values ​​in the reference direction and the cumulative filling amount characterization values ​​in the vertical direction.

[0013] Further, the specific steps for generating the metal hydraulic jet flow rate values ​​for gates in each direction are as follows: Based on the spatial relative deviation, and combined with the preset deviation and widening mapping function, calculate the distribution widening parameter; based on the eccentricity direction angle and the distribution widening parameter, generate the corrected strength mapping curve and construct the flow correction coefficient vector; based on the asymmetric compensation reference value, and combined with the flow correction coefficient vector, perform constraint mapping processing to generate the metal hydraulic jet flow rate values ​​for gates in each direction.

[0014] An infrared positioning-based duct lip casting control system includes: an infrared acquisition unit, used to project an infrared reference spot with a preset geometry into the inner cavity of the duct lip, and to acquire an infrared band image containing the infrared reference spot and the duct lip. The image correction and recognition unit is used to correct the background grayscale non-uniformity of infrared images based on the preset infrared radiation background characteristics of gypsum material, and to extract the actual contour point set of the duct lip and locate the projected geometric center of the infrared reference spot at the sub-pixel level. The positioning adjustment unit is used to perform constrained fitting processing on the actual contour point set using the projected geometric center as the fitting reference, generate a positioning eccentricity vector, and perform positioning adjustment. The infrared flow analysis unit is used to fill the duct lip with molten metal after positioning adjustment, and to slide and collect data from the pouring nozzles in different directions of the duct lip. The corresponding infrared radiation time sequence signal is processed by thermal radiation and flow correlation mapping to generate the filling infrared radiation change rate curve of the gate in each direction; the compensation and deviation analysis unit is used to determine the anisotropic shrinkage characteristics based on the positioning eccentric vector, generate the asymmetric compensation reference value of the gate in each direction, and analyze the spatial relative deviation between the filling infrared radiation change rate curves; the composite mapping flow control unit is used to generate the metal hydraulic jet flow rate value of the gate in each direction based on the spatial relative deviation and the asymmetric compensation reference value, and perform duct lip casting control processing.

[0015] The present invention has the following beneficial effects: (1) The infrared positioning-based duct lip casting control method projects an infrared reference spot into the inner cavity of the lip, collects an infrared band image containing the spot and the lip, and combines the preset infrared radiation background characteristics of the gypsum material to complete the non-uniform grayscale correction of the image background, avoiding the interference of the gypsum mold's own thermal radiation. The Zernike moment sub-pixel edge detection is used to extract the actual contour point set of the lip, and the grayscale centroid method is used to accurately locate the geometric center of the spot projection. Based on this center, a positioning eccentricity vector is generated after coarse error elimination and constraint fitting. The gypsum mold positioning adjustment is completed accordingly, so that the eccentricity offset caused by the drying shrinkage of the gypsum mold can be accurately perceived before mold closing, so as to achieve targeted correction, thereby avoiding the rigid locking of the gypsum mold position and the difficulty in correcting the eccentricity after mold closing, and thus ensuring that the concentricity of the lip mold cavity and the shell main body mold cavity meets the design requirements, and improving the dimensional accuracy of the casting.

[0016] (2) The infrared positioning-based duct lip casting control method, by sliding to collect the infrared radiation time sequence signal of the gate in each direction during the pouring stage, performs first-order difference operation on the signal to obtain the discrete sequence of infrared radiation intensity change rate, inputs the preset radiation and flow rate correlation mapping model, converts it into a filling rate characterization value sequence, and then generates a continuous filling infrared radiation change rate curve through spline interpolation, which intuitively reflects the filling situation of the molten metal in each direction. Based on the positioning eccentric vector, the anisotropic shrinkage characteristics of the plaster mold are extracted to generate the asymmetric compensation benchmark value of the gate in each direction. Combined with the spatial relative deviation of the filling curve, the distribution broadening parameter and flow correction coefficient vector are calculated, and the molten metal hydraulic injection flow rate of each gate is adjusted in real time. This can accurately capture the asymmetric deviation in the filling process and correct the molten metal filling rate in time to avoid defects such as excessive thickness or thinness of the lip due to uneven filling, thereby achieving precise control of the lip wall thickness and improving the casting quality.

[0017] (3) The infrared positioning-based duct lip casting control method completes the precise correction of the plaster mold by infrared positioning before pouring, and monitors the filling status in real time by infrared signal during pouring. Based on the asymmetric reference value of the shrinkage characteristics of the plaster mold and the flow correction based on the spatial deviation, the flow rate of the metal hydraulic injection is dynamically adjusted. Thus, the eccentric positioning correction before pouring, the filling status monitoring and real-time compensation during the pouring process are integrated to form a complete control closed loop, thereby adapting to the shrinkage differences of different individual plaster molds and avoiding product quality fluctuations caused by random distribution of individual shrinkage.

[0018] (4) The infrared positioning-based duct lip casting control system accurately acquires infrared band images containing light spots and duct lips through the infrared acquisition unit. The image correction and recognition unit performs grayscale correction on the image and extracts the contour and locates the center of the light spot to eliminate the interference of the gypsum background and improve the accuracy of feature recognition. The positioning adjustment unit automatically completes the generation of eccentric vector and the positioning adjustment of gypsum mold to improve the concentricity of shell and lip. The infrared flow analysis unit collects the infrared radiation signal of the gate in real time and converts it into the filling change rate curve to provide real-time feedback on the filling status. The compensation and deviation analysis unit combines the positioning eccentric vector to generate compensation reference value and analyze the filling deviation. The composite mapping flow control unit completes the flow calculation and casting control to improve the control accuracy, thereby ensuring the accuracy of detection and adjustment, and making the casting process more stable.

[0019] Of course, any product implementing this invention does not necessarily need to achieve all of the above advantages at the same time. Attached Figure Description

[0020] Figure 1 This is a flowchart of a duct lip casting control method based on infrared positioning according to the present invention. Figure 2 This is a flowchart illustrating the specific steps of background grayscale non-uniformity correction in an infrared band image in a duct lip casting control method based on infrared positioning according to the present invention. Figure 3 This is a block diagram of a duct lip casting control system based on infrared positioning according to the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] The manufacturing of ducted fan housings involves a complete technology chain with multiple coupled processes and progressively increasing precision, encompassing several continuous stages including casting, process inspection, heat treatment control, dimensional finishing, and final performance verification. In this manufacturing chain, the real-time sensing and dynamic control capabilities during the casting filling stage play a decisive role: if the midstream forming stage cannot provide online quantitative feedback on the flow state of the molten metal and the degree of cavity filling, the upstream preset process parameters will become disconnected from the actual physical process, leading to a systematic deviation of the forming result from the design expectations. Therefore, achieving a paradigm shift from "open-loop setting" to "closed-loop control" is a core prerequisite for ensuring the filling quality of complex cavities.

[0023] In the aforementioned technology chain, the lip casting process is a critical node determining the final aerodynamic shape and local wall thickness accuracy. As the leading edge of the ducted fan's intake, the lip's geometric profile and wall thickness distribution directly determine the flow quality and structural load-bearing capacity of the airflow entering the duct. During the injection process, the molten metal enters the mold cavity through the gate; its filling path, solidification sequence, and final distribution are influenced by multiple factors, including mold temperature, injection speed, and mold geometry. The wall thickness uniformity and profile fidelity achieved in this process constitute the core geometric and physical boundaries upon which subsequent manufacturing stages rely.

[0024] However, existing lip casting methods suffer from a bottleneck that restricts the overall accuracy of the process: they typically use preset constant injection parameters for filling, lacking real-time sensing of the actual flow front of the molten metal and the cavity filling status during the gate filling process; when the cavity exhibits eccentricity or uneven wall thickness, the molten metal injection flow rate cannot be dynamically adjusted based on the actual deviation. For thin-walled components with drastic curvature changes, such as duct lips, empirical constant pressure control cannot match the differentiated filling requirements in all directions, leading to local overfilling or underfilling. This control inaccuracy will trigger serious cascading consequences: local wall thickness deviations from the design value at the lip will alter its structural stiffness distribution, causing unexpected reference offsets in subsequent geometric adjustments; profile eccentricity will lead to aerodynamic distortion; and the performance consistency of the final product cannot be guaranteed.

[0025] Based on this, please refer to Figure 1 This invention provides a technical solution: a method for controlling the casting of duct lip based on infrared positioning, comprising the following steps: before the duct lip pouring begins, an infrared reference spot with a preset geometric shape (such as a cross-shaped spot, which facilitates the positioning of the geometric center; the preset geometric center of the infrared reference spot is the positioning reference for the target wall thickness of the lip, corresponding to the theoretical center position of the lip mold cavity) is projected into the inner cavity of the duct lip of the plaster mold to set the positioning reference for the target wall thickness of the lip, and an infrared band image containing the infrared reference spot and the edge of the inner cavity of the duct lip of the plaster mold is acquired; Based on the preset infrared radiation background characteristics of gypsum material, the background gray level of the infrared band image is corrected for non-uniformity, and the actual contour point set of the inner edge of the duct lip is extracted at the sub-pixel level, and the projection geometric center of the infrared reference spot is located. Using the projected geometric center as the fitting reference, the actual contour point set is constrained and fitted to generate a positioning eccentric vector, and the positioning is adjusted. That is, the positioning eccentric vector contains two components, namely the offset in the horizontal direction and the offset in the vertical direction. The base plate currently has a known position in the world coordinate system. Based on the two components of the eccentric vector, the reverse displacement that the base plate needs to move is calculated, and a pulse command is sent to the multi-axis servo positioning mechanism to drive the base plate to move the corresponding compensation distance in the horizontal plane along two orthogonal directions, and continue to move until the actual contour center of the inner cavity of the plaster mold lip coincides with the projected geometric center of the infrared reference spot, so as to ensure the alignment accuracy of the lip mold cavity and the main body mold cavity of the shell and eliminate the eccentric deviation caused by the shrinkage of the plaster mold. After positioning and adjustment, molten metal is poured and filled into the duct lip, and infrared radiation time-series signals corresponding to the gates in different directions of the duct lip cavity are acquired by sliding acquisition (i.e., acquiring signals by continuously intercepting signals in a sliding time window of a preset time length). After thermal radiation and flow correlation mapping processing, the filling infrared radiation change rate curves of the gates in each direction are generated. Based on the positioning eccentric vector, the anisotropic shrinkage characteristics are determined, and the asymmetric compensation reference value of the gates in each direction is generated. The spatial relative deviation between the filling infrared radiation change rate curves of the gates in each direction is analyzed. Based on the spatial relative deviation and asymmetric compensation reference value, the metal hydraulic injection flow rate value of each direction gate (within the current sliding window) is generated through composite mapping processing, and the duct lip casting control processing is performed. That is, the metal hydraulic injection flow rate value is converted into the control command of the die casting machine servo proportional valve and sent to the actuator to adjust the metal flow rate of each direction gate in real time, so as to achieve precise control of duct lip casting and ensure uniform lip wall thickness.

[0026] Specifically, such as Figure 2 As shown, the specific steps for correcting background grayscale non-uniformity in infrared images are as follows: A set of gypsum background calibration images is obtained (dark-field infrared images of the duct lip cavity of each gypsum mold under conditions without infrared reference spot projection are acquired at standard intervals after multiple gypsum molds have dried and before casting, to form a set of gypsum background calibration images), and filtered (such as median filtering) is performed. Specifically, multiple frames of dark-field infrared images under conditions without reference spot are acquired. To improve the signal-to-noise ratio, several frames (e.g., 10 frames) of images are continuously acquired for each gypsum mold and time-domain averaged. The averaged single frame image is used as the calibration image of that gypsum mold. ,in, Indicates the first A set of plaster molds, For the total number of samples (e.g.) (to ensure sufficient sample size), all one sample Together they constitute a plaster background calibration image set , Coordinates in the k-th calibration image The pixel grayscale value at that location; For each set of calibration images in the gypsum background calibration image set For filtering, this embodiment uses median filtering, and the filter window size is selected as follows: pixels, the filtered image is denoted as The filtered plaster background calibration image set is denoted as ; Based on the filtered image set of gypsum background calibration, a surface model (such as a second-order polynomial surface model) characterizing the infrared radiation background distribution of gypsum material at preheated temperature is constructed and stored as a preset infrared radiation background feature of gypsum material. Specifically: All samples Using the grayscale values ​​of corresponding pixel positions as observation data, the least squares method is used to solve for the polynomial coefficients that minimize the overall fitting error, resulting in a second-order polynomial surface model. (Right now The mathematical expression for ) is: ; in, to The model coefficients to be determined (i.e. The optimal combination of coefficients can be obtained by minimizing the sum of squared residuals of all pixels across all samples. The preset infrared radiation background characteristics of gypsum material are solidified and stored in the non-volatile memory of the control system; First, calculate the average gray value at each pixel coordinate, which can be used as the gray value of the gypsum background radiation at that coordinate. The calculation formula is as follows: ; in, pixel coordinates The background radiation grayscale value of the gypsum at that location. The coordinates of the kth group of calibration images in the filtered plaster background calibration image set The pixel grayscale value at that location can further improve the accuracy of the model fitting. The fitting objective is to minimize the mean square error between the model estimate and the actual background radiance grayscale value. The fitting objective function is: ; in, The width of the infrared image is in pixels. The height of the infrared image is represented by the number of pixels. The background estimate of the infrared band image is calculated pixel by pixel based on the surface model, and the background compensation is performed on the infrared band image based on the background estimate to complete the background gray-level non-uniformity correction. Specifically: The infrared image is an infrared image that includes the infrared reference spot and the edge of the cavity. (Right now , Coordinates in the image After considering the pixel grayscale values ​​at the specified location, a second-order polynomial surface model is used. Perform background compensation pixel-by-pixel on the image, that is, for For any pixel location in the model, calculate the baseline estimate of that pixel location using the stored model coefficients. Then, the estimated value is subtracted from the original grayscale value to obtain the corrected image. : ; in, Coordinates in the corrected infrared band image The pixel grayscale value at that location; This is a grayscale compensation constant, with a value between 50 and 100 (to ensure that the grayscale of the corrected image is within a reasonable range, facilitating subsequent edge extraction). To ensure that the grayscale value is non-negative, all values ​​less than zero can be further... Setting it to 0 and completing the above processing achieves the non-uniform correction of background grayscale in infrared band images, effectively suppressing the interference of the plaster mold's own preheating radiation on subsequent edge recognition and spot center positioning.

[0027] The specific steps for extracting the actual contour point set of the inner edge of the duct lip and locating the projected geometric center of the infrared reference spot are as follows: For the infrared band image after background grayscale non-uniformity correction, a sub-pixel edge detection operator based on Zernike moments is used for edge extraction to obtain a set of sub-pixel candidate edge points for the inner edge of the duct lip cavity, specifically: For images Gaussian smoothing is applied to further suppress noise; the standard deviation of Gaussian smoothing is... Set the value to 0.5-1.0, and denote the image after Gaussian smoothing as... ; Calculate the Zernike moment for each pixel in the image, and select... The Zernike moment of order 1, defined as: ; in, for Step Second Zernike moment, The image after Gaussian smoothing is in coordinates The pixel grayscale value at that location, The normalized polar radius of the pixel relative to the center of the detection window; The polar angle of the pixel relative to the center of the detection window; for Step Zernike polynomials of degree Zernike; For indicator functions, when hour, ,otherwise Used to define the range of the detection window (the detection window is preferably...) window); Based on the amplitude and phase of the Zernike moment, it is determined whether a pixel is an edge point. When the amplitude of the Zernike moment exceeds a preset edge threshold (the threshold can be set to 50-80), the pixel is determined to be an edge point. At the same time, the sub-pixel coordinates of the edge points are calculated by using the phase of the Zernike moment, thus obtaining a sub-pixel precision candidate edge point set for the inner edge of the duct lip. (Right now ),in, The index of the candidate edge point. For the first Sub-pixel coordinates of candidate edge points; Connectivity analysis is performed on the candidate edge point set, and morphological closing operations are performed based on the results of the connectivity analysis to obtain the actual contour point set, specifically as follows: The connected component analysis employs an 8-neighborhood search algorithm to group spatially adjacent candidate edge points into the same connected component, setting a threshold for the number of connected points. (For example, take) 20% of all points containing less than Isolated, discrete edge segments are treated as pseudo-edges or noise and removed, while continuous edge point connected regions are preserved; Morphological closing operations are performed on the preserved main connected components. The closing operation first dilates the binarized edge image to bridge narrow gaps that are broken due to low contrast, and then performs erosion to restore the original width of the edges. The size of the structuring element for the closing operation can be selected according to the image resolution. For example, a disk-shaped structuring element with a radius of 3 pixels (or a 3×3 structuring element) can be used to obtain the actual contour point set that can accurately reflect the true contour of the inner cavity of the plaster mold lip. ,in, The index of the actual contour point. For the first Subpixel coordinates of each actual contour point; Adaptive threshold segmentation is performed on the region containing the infrared reference spot in the infrared band image to extract the initial region of the spot. The square of the gray value of each pixel in the initial region of the spot is used as the weight, and the first moment and zero moment of the initial region of the spot are calculated using the gray-level centroid method. The projected geometric center of the infrared reference spot is then located based on these moments. Specifically: Infrared reference spot in image The area appears as a bright, approximately circular region (the grayscale distribution of the cross-shaped spot still conforms to the characteristic of a bright center and dark edges). First, adaptive threshold segmentation is performed on this region, for example, using the Otsu method (maximum inter-class variance method) to automatically determine a segmentation threshold. To convert images with gray values ​​greater than 10 ... The pixels are initially classified as light spot areas. (The gray value of the spot area is higher than the segmentation threshold, while the gray value of the background area is lower than the segmentation threshold.) For the initial region of the light spot Each pixel within Let its grayscale value be Calculate the zeroth moment of this region. and first moment , : ; ; ; in, The sum of grayscale weights of the light spot region (i.e. This reflects the overall gray intensity of the light spot; for Gray-scale weighting moments in the direction (i.e.) ), for Gray-scale weighting moments in the direction (i.e.) ), used to characterize the center position of the grayscale distribution of the light spot; The sub-pixel level projection geometric center coordinates of the infrared reference spot for: This coordinate represents the precise projection position of the infrared reference spot onto the image plane.

[0028] The specific steps for generating the positioning eccentricity vector are as follows: Using the projected geometric center as the initial fitting circle center, coarse error removal is performed on the actual contour point set of the inner cavity. The 3σ criterion can be used to filter out abnormal edge points whose distance from the initial fitting circle center exceeds a preset deviation threshold, resulting in the filtered contour point set, as follows: The 3σ criterion (Laida criterion) is used for screening, and the contour point set is calculated. All pixels in the circle to the initial center distance The calculation formula is: ; in, For the first The distance from each actual contour point to the center of the initial fitted circle; And calculate the mean of the distance sequence. (Right now ) and standard deviation This is used to set the deviation threshold coefficient. (usually taken) ), to satisfy all (Right now or ) points Points identified as abnormal edges (caused by image noise, edge extraction error, or local defects in plaster mold) are removed. (It should be noted that if the actual collected contour point distance distribution shows a significant skewness or there are many outliers, a robust screening method based on the absolute deviation of the median can be used as an alternative, i.e., the median of the distance sequence from all contour points to the initial fitted circle center is calculated and denoted as the center distance.) The absolute deviation between the distance to each contour point and the center distance is calculated, and the median of these absolute deviations is used as a measure of the dispersion of the distance distribution. For any contour point, if the absolute value of the difference between its distance and the center distance exceeds the product of a preset threshold coefficient and the above dispersion measure, then the point is identified as an abnormal edge point and is removed. Maintain distance (Right now The contour points within the range constitute the filtered contour point set. ,in, , This represents the number of points in the filtered outline. Based on the filtered contour point set and combined with preset constraints, a constrained least squares circle fitting objective function is established, and the optimal circle center coordinates and optimal radius values ​​are generated by solving the problem. Specifically: The constraint is to limit the fitting radius. The allowable variation range corresponds to the statistical deviation range of irregular shrinkage deformation of the plaster mold, which can be set to the nominal radius. (Right now The theoretical design radius of the duct lip cavity is determined based on the product drawings. (Converted to pixel units), the objective function and constraints are as follows: ; ; in, Let be the theoretical pixel radius of the duct lip cavity. The allowable radius deviation threshold (corresponding to ±0.2mm); To solve for the optimal center coordinates and radius under constraints Using the Lagrange multiplier method, the inequality constraint is transformed into an equality constraint and then introduced into the objective function to form an unconstrained augmented objective function, thus constructing the Lagrange function:

[0029] ; in, , To presuppose Lagrange multipliers, , When the constraint is not activated (i.e. ), When the constraint is activated (i.e. or ), corresponding to non-zero Lagrange multipliers; For the Lagrange function To each , , , , Find the partial derivatives and set them equal to zero to obtain a system of equations. Solve this system of equations using an iterative optimization algorithm (such as the Levenberg-Marquardt algorithm) to obtain the optimal coordinates of the circle center under the constraints. and optimal radius value The iteration termination condition is that the deviation of the center coordinates between two iterations is less than 0.001mm and the radius deviation is less than 0.001mm, ensuring that the fitting accuracy meets the positioning adjustment requirements; The optimal center coordinates and the projected geometric center are subtracted by vector to generate a positioning eccentricity vector, specifically as follows: For the optimal center coordinates With the geometric center of projection Perform vector subtraction to generate a positioning eccentricity vector. The calculation is as follows: ; in, To locate the eccentric vector in Component of direction; To locate the eccentric vector in The directional component.

[0030] In this implementation scheme, background calibration and filtering can effectively suppress radiation interference generated by plaster mold preheating, ensuring clear and usable infrared images; subpixel edge detection and morphological processing can accurately extract the actual contour of the inner cavity of the lip, eliminate noise interference, and ensure contour integrity; grayscale centroid method can accurately locate the center of the infrared reference spot, and error elimination and constraint fitting can accurately generate the positioning eccentric vector, providing a reliable basis for plaster mold position adjustment, thereby ensuring uniform casting size and improving the overall casting effect and stability.

[0031] Specifically, the steps for generating the rate of change curves of infrared radiation of the gate in each direction are as follows: For the infrared radiation time series signal, first-order difference operations are performed to obtain discrete sequences of the rate of change of infrared radiation intensity at the gate in each direction, as follows: Let the first Gating gates in one direction ( , (The total number of gates, evenly distributed around the lip) within the current sliding time window The infrared radiation time series signal collected internally is a time series. ( Sampling time, Let Q be the total number of sampling points within the sliding time window at the q-th sampling time. For the first Each gate at time (infrared radiation intensity value), where At the sampling time, to eliminate high-frequency random noise in the signal, we can first... Perform moving average filtering with a window width of 5 to 10 sampling points; Performing a first-order difference operation on the filtered sequence yields a discrete sequence of the rate of change of infrared radiation intensity at the gate in that direction. : ; In the formula, The rate at which infrared radiation intensity changes over time is characterized by the faster the molten metal filling rate; the more drastic the change in infrared radiation intensity, the greater the rate of change. The sampling period is set to 0.01-0.05s. , For the u-th gate, , (infrared radiation intensity value at time). This represents the number of sampling points within the window. The discrete sequence of infrared radiation intensity change rate is input into a preset radiation-flow correlation mapping model, and radiation-flow transformation processing is performed to output a sequence of filling rate characterization values ​​for gates in each direction (i.e., the filling rate characterization values ​​of gates in each direction at each sampling time within the current sliding time window), specifically: Discretized sequence of infrared radiation intensity change rate Input a preset radiation-flow correlation mapping model This model is a nonlinear mapping relationship established in advance through numerous calibration experiments. Its function is to convert the rate of change of thermal radiation sensed by the infrared detector into the instantaneous filling rate of the molten metal at the corresponding gate. For the first... The model outputs a sequence of fill rate representation values ​​at each sampling time within the current sliding time window for each of the three gate directions. : ; in, To and The corresponding characterization value of the molten metal filling rate; For the current sliding time window, the first The sequence of filling rate characterization values ​​for each gate is as follows: This sequence can reflect the current time window in real time. The variation of the molten metal filling rate at each gate; The filling rate characterization value sequence is processed by spline interpolation (such as cubic spline interpolation) to generate the filling infrared radiation change rate curves of the gates in each direction, as follows: Cubic spline interpolation is used, which ensures the continuity of the second derivative at the nodes while exhibiting good smoothness for time intervals. , in time As a node, with Construct cubic spline functions for the nodal function values. The function satisfies the following conditions: In each sub-interval superior, It is a cubic polynomial; ; Throughout the entire range It has continuous first and second derivatives; The boundary conditions adopt natural boundary conditions, that is, the second derivative at the endpoints is zero: , ; Based on the above constraints, by solving the system of linear equations composed of nodal function values, first-order derivative continuity, and second-order derivative continuity, the coefficients of the cubic polynomial in each subinterval can be uniquely determined. Thus, the discrete filling rate characterization sequence, originally defined only at each sampling time, is analytically extended into a smooth curve that is continuous in time and second-differentiable, i.e., the filling infrared radiation change rate curves of the gates in each direction. It can intuitively reflect the real-time changes in the filling rate of molten metal in all directions.

[0032] The preset steps of the radiation-flow correlation mapping model are as follows: Multiple sets of duct lip samples were collected with time-series infrared radiation signals from the gating gate under different metal hydraulic injection flow rates, and the actual metal hydraulic injection flow rate curves were recorded simultaneously to form a radiation and flow rate calibration dataset, which is as follows: Under standard casting conditions (standard casting conditions can be: molten metal temperature of 1500-1550℃, plaster mold preheating temperature of 80-100℃, and basic molten metal hydraulic injection flow rate of 8-12L / min), multiple sets of duct lip samples were collected with gate infrared radiation time-series signals under different molten metal hydraulic injection flow rates. The adjustment range of the molten metal hydraulic injection flow rate was set to 50-200mm / s (or 0.5m / s~3.0m / s). 10-15 different molten metal hydraulic injection flow rate levels were selected (evenly distributed within the adjustment range, such as low speed 0.5m / s, medium speed 1.5m / s, high speed 3.0m / s, etc.), and each level was repeated 3-5 times. For each experiment, infrared radiation timing signals were collected from each gate. Simultaneously, the actual displacement and time curve of the injection punch were recorded. Then, by differentiation, the actual metal hydraulic jet flow rate curve can be obtained. The calculation formula is: ; in, This represents the actual hydraulic injection flow rate of the molten metal. The first derivative of the displacement signal; Infrared radiation time-series signals from all experiments Corresponding actual metal hydraulic injection flow rate curve Together, they constitute the radiation and flow calibration dataset. ; The radiation and flow calibration dataset is divided into windows, and the labels for the infrared radiation intensity change rate samples and flow samples are calculated for the sliding time window. Specifically: Radiation and flow calibration dataset Windowing is performed, and the infrared radiation intensity change rate sample and flow rate sample label are calculated for each sliding time window. The windowing uses the same parameters as described above. For each group of infrared radiation time series signals in the calibration dataset... Iterate through the entire timeline using a sliding time window. For any sliding time window, perform the following operations: Calculate the first-order difference mean of the infrared radiation intensity within the window, and use it as a sample of the rate of change of infrared radiation intensity within the window. ; The time integral average of the actual metal hydraulic injection flow rate within the window is calculated and used as the flow rate sample label for that window. ; The above processing was performed on all sliding time windows of all experimental groups, and a large amount of data was collected. Sample pairs constitute the model training dataset. ,in The total number of sample pairs; Using the rate of change of infrared radiation intensity in each sliding time window as the independent variable and the flow rate sample label as the dependent variable, a radiation-flow correlation mapping model is established using a data-driven nonlinear regression method. Specifically: As a preferred implementation method, the radiation-flow correlation mapping model can be established using a hybrid regression method combining quadratic polynomial regression and radial basis kernel functions, the specific form of which is as follows: ; In the formula, The rate of change of infrared radiation intensity is the input (independent variable). These are the coefficients of a quadratic polynomial; The number of support vectors (adaptively determined by the training dataset); For the first The weight coefficients of each support vector; This is the bandwidth parameter of the radial basis function (which controls the range of influence of the kernel function); For the first Samples of infrared radiation intensity change rates corresponding to each support vector; This is the radial basis kernel function, used to fit nonlinear mapping relationships; It should be noted that support vectors are derived from the model training dataset. In this process, key samples are selected by solving an objective function (minimizing the mean squared error between the predicted value and the sample label, plus regularization constraints). Specifically, during training, not all... Corresponding sample pairs All factors will affect the final mapping relationship of the model. Only samples that support the nonlinear relationship between the rate of change of infrared radiation and the flow rate of molten metal (i.e., significantly affect the shape of the fitted curve) will be selected as support vectors. The remaining samples (non-support vectors) have no impact on the solution of model parameters and can be ignored. The selection process is automatically completed by the gradient descent algorithm when solving the objective function, without manual intervention. The final number of selected support vectors is [the number of vectors selected]. ; The goal of model training is to minimize the mean squared error (MSE) between the predicted values ​​and the sample labels. The objective function is: ; in, This is the regularization coefficient (with a value of 0.001-0.01), used to prevent overfitting and balance fitting accuracy with model complexity. This is the mean squared error term, which reflects the fitting accuracy of the model; This is a regularization term used to constrain the magnitude of model parameters and prevent the model from overfitting due to excessively large parameters. The optimal structural parameters of the radiation-flow correlation mapping model are determined using cross-validation and parameter optimization strategies, and the trained model parameters are then stored in a fixed manner. Specifically: A cross-validation and parameter optimization strategy is used to determine the radiation-flow correlation mapping model. The optimal structure parameters for the dataset The dataset is randomly divided into a training set (70%), a validation set (20%), and a test set (10%). Train the model on the training set and evaluate different combinations of hyperparameters (such as regularization coefficients) on the validation set. Radial basis kernel function bandwidth The model performance under these conditions was assessed by using a grid search method to traverse combinations of hyperparameters (such as...). Take values ​​of 0.001, 0.005, and 0.01; Choose the hyperparameter combination that minimizes the root mean square error (RMSE) on the validation set (0.1, 0.5, 1.0, 2.0) as the optimal parameter combination. The generalization ability of the final model is evaluated on the test set, such as RMSE ≤ 0.5 mm / s on the test set, to ensure that the model can accurately map the relationship between the rate of change of infrared radiation and the actual flow rate of the molten metal. The trained model parameters ( The data is stored in the memory of the control system, forming a radiation-flow correlation mapping model.

[0033] In this implementation scheme, by filtering and differentially processing the infrared radiation time-series signal, noise interference can be effectively eliminated, and the changing trend of infrared radiation at each gate can be accurately captured. Combined with the preset mapping model, the infrared radiation change rate can be accurately converted into the molten metal filling rate. Furthermore, spline interpolation processing can transform discrete rate data into continuous and smooth curves to present the filling dynamics of gates in each direction. Secondly, the mapping model is constructed through multiple sets of calibration experiments. Combined with nonlinear regression and parameter optimization, it can accurately match the relationship between radiation change and actual flow rate, avoiding mapping deviation. This avoids the problem that it is difficult to capture the radiation signal and accurately correspond the actual flow rate in real time, thus ensuring precise control of the lip wall thickness and improving the casting quality.

[0034] Specifically, the steps for generating the asymmetric compensation reference values ​​for gates in each direction are as follows: The positioning eccentricity vector is decomposed to extract anisotropic shrinkage characteristics, including the eccentricity direction angle and eccentricity distance. The direction of the gate with the maximum resistance corresponding to the eccentricity direction angle is determined, and the gate with the maximum resistance direction is used as the reference compensation direction. Specifically: For the positioning eccentric vector Perform vector decomposition to extract anisotropic contraction features. The polar coordinate representation is: eccentricity and eccentricity direction angle These two parameters fully describe the shrinkage characteristics of this plaster mold individual: the main direction of shrinkage is... The direction indicated, the shrinkage strength and The magnitude is positively correlated ( The larger the size, the more intense the contraction. Eccentricity direction angle The direction it points to is determined to be the direction of the gate with the greatest resistance, and from... Find the corresponding gate in each direction. The gate with the closest direction is assigned a number. , as the reference compensation direction (i.e., the gate corresponding to the main shrinkage direction); The installation orientation angle of each gate is preset to be... , uniformly distributed in [0, Within the interval (e.g.) hour, , , , The gate corresponding to the reference compensation direction satisfies: That is, the installation direction angle and the eccentricity direction angle of the gate. The difference is the smallest; Taking the reference compensation direction as the center and combining it with the eccentricity, the compensation weight coefficient of the gate in each direction is calculated as follows: For the u-th direction gate, calculate its installation direction angle. Compensation direction with reference ( The angle difference between The calculation formula is: ; Will Normalization to interval (if) Then take ), and the reference compensation direction corresponds to (Right now hour, ); Calculate the foundation weight corresponding to this gate. A cosine squared decay function is used to ensure maximum weight in the reference direction and a smooth decrease towards both sides. The formula is as follows: ; in, [0, 1], when (In the direction of reference compensation) (Highest weight); when When (opposite to the reference direction), (Minimum weight) conforms to the gradient distribution characteristics of plaster mold shrinkage; Introducing (normalized) eccentricity The concentration factor is used to adjust the concentration of the weight distribution. A larger eccentricity means that the shrinkage is more concentrated in the main direction, so the compensation should also be more concentrated at the reference gate; a smaller eccentricity means that the shrinkage is more uniform, and the compensation weight distribution should be more gradual. It is a kind of The positive correlation function has the following formula: ; in, The preset scaling factor (values ​​range from 0.5 to 2.0); when (Without bias) The weight distribution maintains a smooth state similar to the basic weights; when hour, The weight distribution curve becomes sharper, and the energy is more concentrated in direction; By using basic weights Perform an exponential transformation to obtain the final compensation weight coefficients. The formula is: ; It should be noted that the weight of the benchmark compensation direction... The reference direction remains constant to ensure maximum compensation; the weights of other gates vary. The value decreases as the value increases, and the magnitude of the decrease increases with the value of the value. The increase in the value exacerbates the problem, achieving a precise match between the compensation weight and the shrinkage characteristics of the plaster mold; The compensation weighting coefficients are corrected against the preset base metal hydraulic injection flow rate to obtain the asymmetric compensation reference values ​​for the gates in each direction, specifically: For the Each gate has a directional gate, and its asymmetric compensation reference value The calculation is as follows: ; in, This is the weighting normalization coefficient, used to ensure that the sum of the compensation reference values ​​for all gates equals the total flow rate under the unbiased state. To ensure consistency and avoid deviations in total fill volume due to weight allocation.

[0035] The specific steps for analyzing the spatial relative deviation between the infrared radiation change rate curves of the gates in different directions are as follows: By integrating the rate of change of infrared radiation of the filling at each gate direction over time, a spatial filling distribution vector is constructed, which is as follows: For the current sliding time window (corresponding time interval) ), for each direction of the gate Calculate the definite integral of its rate of change curve within the window to obtain the instantaneous cumulative filling amount characterization value of the gate in that direction within the time window. The integral formula is: ; in, For the first The cumulative filling amount of each gate within the current time window; , These are the start and end times of the current sliding time window, respectively. For the first The rate of change of infrared radiation during filling of each gate (i.e., the curve of the change of the metal molten filling rate characterization value over time). All Gates in each direction The values ​​are collected to form a space-filling distribution vector. This vector reflects the progress or activity of the molten metal filling in different directions at the current instant. The larger the value, the faster the molten metal filling rate and the greater the filling volume in that direction; conversely, the smaller the value, the greater the filling rate and the greater the filling volume of the molten metal in that direction. The smaller the value, the slower the filling rate and the less filling volume in that direction; Using the gate direction pointed to by the eccentricity angle as the reference direction, the elements in the spatial filling distribution vector are sorted according to the gate direction to form a sequence of direction and cumulative filling amount, specifically as follows: The elements in the space-filling distribution vector are arranged according to the gate direction angle. The sizes are reordered to form a sequence of direction and cumulative filling volume, assuming the gate direction angle. From 0 to If arranged in ascending order, the sorted sequence is: ; From the direction and cumulative fill volume sequence, extract the cumulative fill volume characterization value in the reference direction and the cumulative fill volume characterization value in the direction perpendicular to the reference direction; based on the cumulative fill volume characterization values ​​in the reference direction and the cumulative fill volume characterization values ​​in the perpendicular direction, generate the spatial relative deviation (i.e., the degree to which the ratio of the cumulative fill volume characterization value in the direction perpendicular to the eccentric direction angle deviates from the ideal uniform value of 1), specifically: Extract the baseline direction (i.e., ...) from the direction and cumulative fill volume sequence. Cumulative fill volume characterization value in the direction) : If the reference compensation direction If it corresponds exactly to an element in the sorted sequence, then If the reference direction does not have a gate (i.e. Not with any (If they overlap), then through the two adjacent gates The value is obtained by linear interpolation. perpendicular to the reference direction (i.e.) Cumulative fill volume characterization value in the direction) : Adopted and Extract using the same method; if there is a gate in the vertical direction, then... For this gate Value; if not, then through the adjacent gate. Values ​​are obtained through linear interpolation; Calculate spatial relative deviation It is the cumulative filling amount in the vertical direction of the sequence. Cumulative filling amount in the reference direction The degree to which the ratio deviates from the ideal uniform value of 1 is calculated using the following formula: .

[0036] The specific steps for generating the metal hydraulic injection flow rate values ​​for gates in various directions are as follows: Based on the spatial relative deviation, and combined with the preset deviation and broadening mapping function, the distribution broadening parameter is calculated, specifically as follows: The mapping function between deviation and broadening is specifically formulated as follows: ; in, is the distribution broadening parameter (dimensionless); The maximum expansion parameter (corresponding to) When the filling is completely uniform, the value is 0.5-1.0. The attenuation coefficient is 5-10 (the broadening parameter controls the attenuation). (the rate of decay); Based on the eccentricity direction angle and distribution broadening parameters, a corrected strength mapping curve distributed along the gate direction is generated, and a flow correction coefficient vector is constructed, specifically as follows: Eccentricity direction angle The direction of the gate (i.e., the aforementioned reference compensation direction) The peak direction of this compensation mapping is determined, and its index is recorded. This direction requires the largest possible feedback correction to offset the greatest flow resistance; by Centered on, with As the attenuation control factor, a corrected strength mapping curve is generated. This curve reflects the feedback correction strength of each gate. For any given... For gates in each direction, the corrected strength Calculate using the following formula: ; in, For the first The corrected strength (dimensionless) of each gate, with a value range of [value missing]. ; To correct the upper limit of the amplitude, it can be set to 1 to control the maximum range of the correction intensity; ( ), for the first The direction angle difference between each gate and the peak direction is normalized to... arrive Within the range; This is a Gaussian attenuation term, which controls the rate at which the correction intensity decays with the difference in direction angle. The smaller the value, the faster the decay, and the more concentrated the correction energy is in the peak direction; For a sign function, when When (i.e., the gate is located within ±90° of the peak direction), A positive correction (increased flow) is obtained; when When (i.e., the gate is located within the range of 90°-270° in the peak direction), To obtain negative correction (reduce flow); when hour, No corrections; Corrected strength of gates in all directions Arranged sequentially according to gate number u=1, 2, ..., U, forming a flow correction coefficient vector. This vector reflects the direction and intensity of the feedback correction required for each gate, positive... This indicates that the flow rate of the metal hydraulic injection at this gate needs to be increased; a negative flow rate indicates a need for increased flow rate. This indicates that the flow rate of the metal hydraulic injection at this gate needs to be reduced; the larger the absolute value, the greater the correction range. Based on the asymmetric compensation reference value and combined with the flow correction coefficient vector for constraint mapping processing, the metal hydraulic jet flow rate values ​​for each direction of the gate are generated, specifically as follows: Set feedback correction gain coefficient It is used to adjust the overall strength of feedback correction, avoiding over- or under-correction. The value ranges from 0.1 to 0.3, and its magnitude corresponds to the spatial relative deviation. A positive correlation can be dynamically adjusted through the following linear relationship: ,when hour, (Maximum correction strength); when hour, (Minimum correction strength) ensures that the correction strength is precisely matched with the degree of fill asymmetry; Calculate the feedback correction flow rate for each gate. The formula is: ; in, The feedback correction flow rate for the u-th gate (unit: L / min); To correct the gain coefficient for feedback; The corrected strength for the u-th gate; This is the asymmetric compensation benchmark value for the u-th gate; Generate the final metal hydraulic injection flow rate values ​​for gates in each direction. The calculation formula is: ; In the formula, The final hydraulic injection flow rate setting for the u-th gate must meet the following process constraints: ,in, Minimum permissible metal hydraulic injection flow rate (e.g., 5-7). The maximum allowable flow rate of the metal hydraulic injection is set to 15-18. If the calculated result exceeds the constraint range, the corresponding boundary value is taken as the final flow rate value.

[0037] In this implementation scheme, by decomposing the positioning eccentricity vector, shrinkage characteristics are extracted and the benchmark compensation direction is determined. Combined with the eccentricity, the compensation weight is calculated, and the generated asymmetric compensation benchmark value can accurately match the individual shrinkage differences of the plaster mold. Furthermore, by integral analysis of the filling infrared radiation change rate curve, the filling deviation in each direction can be accurately captured. The generated spatial relative deviation can reflect the uniformity of filling. Finally, based on the deviation and the compensation benchmark value, the hydraulic injection flow rate of metal at each gate is generated through correction mapping to achieve precise flow rate control. This compensates for uneven shrinkage and filling deviation, ensures uniform filling of molten metal, reduces defects such as uneven casting wall thickness, and thus ensures the accuracy of casting control.

[0038] Please see Figure 3 This invention provides a technical solution: a duct lip casting control system based on infrared positioning, comprising: an infrared acquisition unit for projecting an infrared reference spot with a preset geometric shape into the inner cavity of the duct lip and acquiring an infrared band image including the infrared reference spot and the duct lip; an image correction and recognition unit for performing background grayscale non-uniform correction on the infrared band image based on preset infrared radiation background characteristics of gypsum material, and extracting the actual contour point set of the duct lip and locating the projection geometric center of the infrared reference spot at the sub-pixel level; and a positioning adjustment unit for constraining and fitting the actual contour point set with the projection geometric center as the fitting reference, generating a positioning eccentricity vector, and performing positioning adjustment. The infrared flow analysis unit is used to fill the duct lip with molten metal after positioning and adjustment, and to collect the infrared radiation time sequence signals corresponding to the gates in different directions of the duct lip. After thermal radiation and flow correlation mapping processing, the filling infrared radiation change rate curves of the gates in each direction are generated. The compensation and deviation analysis unit is used to determine the anisotropic shrinkage characteristics based on the positioning eccentricity vector, generate the asymmetric compensation reference value of the gates in each direction, and analyze the spatial relative deviation between the filling infrared radiation change rate curves. The composite mapping flow control unit is used to generate the molten metal hydraulic injection flow rate value of the gates in each direction based on the spatial relative deviation and the asymmetric compensation reference value, and to perform casting control processing of the duct lip.

[0039] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0040] 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 controlling the casting of duct lip based on infrared positioning, characterized in that, Includes the following steps: An infrared reference spot with a preset geometric shape is projected into the inner cavity of the duct lip, and an infrared band image containing the infrared reference spot and the duct lip is acquired. Based on the preset infrared radiation background characteristics of gypsum material, the infrared band image is subjected to background grayscale non-uniformity correction, and the actual contour point set of the duct lip is extracted at the sub-pixel level and the projection geometric center of the infrared reference spot is located. Using the projected geometric center as the fitting reference, the actual contour point set is subjected to constraint fitting to generate a positioning eccentricity vector, and positioning adjustment is performed. After positioning and adjustment, molten metal is poured and filled into the duct lip, and infrared radiation time-series signals corresponding to the gates in different directions of the duct lip are collected by sliding. After thermal radiation and flow correlation mapping processing, the filling infrared radiation change rate curves of the gates in each direction are generated. Based on the positioning eccentric vector, the anisotropic shrinkage characteristics are determined, and the asymmetric compensation reference values ​​of the gates in each direction are generated. The spatial relative deviation between the filling infrared radiation change rate curves is also analyzed. Based on the spatial relative deviation and the asymmetric compensation reference value, the metal hydraulic injection flow rate value of the gate in each direction is generated through composite mapping processing, and the casting control processing of the duct lip is performed.

2. The duct lip casting control method based on infrared positioning according to claim 1, characterized in that, The specific steps for correcting the background grayscale non-uniformity of the infrared band image are as follows: Obtain a set of gypsum background calibration images and perform filtering processing; Based on the filtered gypsum background calibration image set, a curved surface model is constructed and stored as a preset infrared radiation background feature of gypsum material; Based on the surface model, the background estimate is calculated pixel by pixel for the infrared band image, and background compensation is performed on the infrared band image based on the background estimate to complete the background grayscale non-uniformity correction.

3. The duct lip casting control method based on infrared positioning according to claim 2, characterized in that, The specific steps for extracting the actual contour point set of the inner edge of the duct lip and locating the projected geometric center of the infrared reference spot are as follows: For the infrared band image after background grayscale non-uniformity correction, an edge extraction is performed using a sub-pixel edge detection operator based on Zernike moments to obtain a candidate edge point set for the inner edge of the duct lip cavity; Connectivity analysis is performed on the candidate edge point set, and morphological closing operation is performed based on the connected component analysis results to obtain the actual contour point set of the inner edge of the duct lip. Adaptive threshold segmentation is performed on the region where the infrared reference spot is located in the infrared band image to extract the initial region of the spot. The first moment and zero moment of the initial region of the spot are calculated using the gray-scale centroid method, and the projection geometric center of the infrared reference spot is located.

4. The duct lip casting control method based on infrared positioning according to claim 3, characterized in that, The specific steps for generating the positioning eccentricity vector are as follows: Using the projection geometric center as the initial fitting circle center, the actual contour point set of the inner cavity is subjected to coarse error elimination to obtain the filtered contour point set. Based on the filtered contour point set and combined with the preset constraints, a constrained least squares circle fitting objective function is established, and the optimal circle center coordinates are generated by solving the problem. The optimal circle center coordinates and the projected geometric center are vector subtracted to generate a positioning eccentricity vector.

5. The duct lip casting control method based on infrared positioning according to claim 1, characterized in that, The specific steps for generating the rate of change curves of infrared radiation of the gate in each direction are as follows: The infrared radiation time-series signal is subjected to first-order difference operation to obtain discrete sequences of the rate of change of infrared radiation intensity of the gate in each direction. The discrete sequence of infrared radiation intensity change rate is input into a preset radiation and flow rate correlation mapping model, and radiation and flow rate transformation processing is performed to output a sequence of filling rate characterization values ​​for gates in each direction. Spline interpolation is performed on the filling rate characterization value sequence to generate filling infrared radiation change rate curves for gates in each direction.

6. The duct lip casting control method based on infrared positioning according to claim 5, characterized in that, The preset steps of the radiation-flow correlation mapping model are as follows: Multiple sets of duct lip samples were collected with time-series infrared radiation signals of the gate under different metal hydraulic jet flow rates, and the actual metal hydraulic jet flow rate curves were recorded simultaneously to form a radiation and flow rate calibration dataset. The radiation and flow calibration dataset is divided into windows, and the labels for the infrared radiation intensity change rate samples and flow samples are calculated. Using the infrared radiation intensity change rate sample as the independent variable and the flow rate sample label as the dependent variable, a radiation-flow correlation mapping model is established using a data-driven nonlinear regression method. The optimal structural parameters of the radiation-flow correlation mapping model are determined by cross-validation and parameter optimization strategies.

7. The duct lip casting control method based on infrared positioning according to claim 1, characterized in that, The specific steps for generating the asymmetric compensation reference values ​​for gates in various directions are as follows: The positioning eccentric vector is decomposed to extract anisotropic shrinkage features, including eccentric direction angle and eccentricity, and the direction of the gate with the maximum resistance corresponding to the eccentric direction angle is determined. The gate with the maximum resistance is used as the reference compensation direction. The compensation weight coefficients for each direction of the gate are calculated using the reference compensation direction as the center and in combination with the eccentricity. The compensation weight coefficients are corrected by comparing them with the preset base metal hydraulic injection flow rate to obtain the asymmetric compensation reference values ​​for the gates in each direction.

8. The duct lip casting control method based on infrared positioning according to claim 7, characterized in that, The specific steps for analyzing the spatial relative deviation between the infrared radiation change rate curves of the gates in different directions are as follows: By performing time integration on the aforementioned rate of change curves of infrared radiation, a spatial filling distribution vector is constructed. Using the gate direction pointed to by the eccentric direction angle as the reference direction, the spatial filling distribution vector is sorted according to the gate direction to form a sequence of direction and cumulative filling amount; From the direction and cumulative fill amount sequence, extract the cumulative fill amount characterization value in the reference direction and the cumulative fill amount characterization value in the direction perpendicular to the reference direction; The spatial relative deviation is generated based on the cumulative fill volume characterization value in the reference direction and the cumulative fill volume characterization value in the vertical direction.

9. The duct lip casting control method based on infrared positioning according to claim 8, characterized in that, The specific steps for generating the metal hydraulic injection flow rate values ​​for gates in various directions are as follows: Based on the spatial relative deviation, and combined with the preset deviation and widening mapping function, the distribution widening parameter is calculated; Based on the eccentricity direction angle and the distribution broadening parameters, a corrected intensity mapping curve is generated, and a flow correction coefficient vector is constructed. Based on the aforementioned asymmetric compensation reference value, and combined with the flow correction coefficient vector for constraint mapping processing, the metal hydraulic jet flow rate values ​​for each direction of the gate are generated.

10. A duct lip casting control system based on infrared positioning, employing the duct lip casting control method based on infrared positioning as described in any one of claims 1-9, characterized in that, The system includes: An infrared acquisition unit is used to project an infrared reference spot with a preset geometric shape into the inner cavity of the duct lip, and to acquire an infrared band image containing the infrared reference spot and the duct lip. The image correction and recognition unit is used to perform background grayscale non-uniformity correction on the infrared band image based on the preset infrared radiation background characteristics of gypsum material, and to extract the actual contour point set of the duct lip and locate the projection geometric center of the infrared reference spot in subpixel extraction. The positioning adjustment unit is used to perform constraint fitting processing on the actual contour point set with the projected geometric center as the fitting reference, generate a positioning eccentricity vector, and perform positioning adjustment. The infrared flow analysis unit is used to fill the duct lip with molten metal after positioning and adjustment, and to slide and collect the infrared radiation time sequence signals corresponding to the gates in different directions of the duct lip. After thermal radiation and flow correlation mapping processing, the filling infrared radiation change rate curves of the gates in each direction are generated. The compensation and deviation analysis unit is used to determine the anisotropic shrinkage characteristics based on the positioning eccentric vector, generate the asymmetric compensation reference value of the gate in each direction, and analyze the spatial relative deviation between the filling infrared radiation change rate curves. The composite mapping flow control unit is used to generate the metal hydraulic injection flow rate value of the gate in each direction based on the spatial relative deviation degree and the asymmetric compensation reference value, and to perform duct lip casting control processing.

Images