Visual inspection method for defects of injection-molded interior parts and self-adaptive regulation method for process parameters

Abstract

The present application relates to the field of intelligent manufacturing, and specifically discloses a method for visual detection of injection-molded interior part defects and self-adaptive regulation of process parameters, which comprises: collecting a pressure response sequence at a weld line, separating impact energy and relaxation decay rate, and comparing with ideal values to generate a weld weakness index distribution; secondly, delimiting a valley visual shielding area by using a texture drawing and an illumination model, positioning a hidden defect candidate area through spatial position registration; collecting a developed image and performing gradient direction filtering to enhance strip features, removing texture interference through morphological deviation analysis of direction, length, width, etc., and confirming real defects; finally, extracting the decay degree at the defect, tracing corresponding mold temperature control units and injection speed sections, and generating temperature and speed compensation instructions using a response surface model; the method realizes confirmation and closed-loop regulation of visual blind area weld defects, avoids missing judgment of strength hidden dangers or blindly adjusting parameters to damage the texture appearance.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing, specifically to a method for visual inspection of defects in injection-molded interior parts and adaptive control of process parameters. Background Technology

[0002] In the injection molding of imitation leather or wood grain trim panels for automotive interiors, weld lines, as weak joints formed by the convergence of melt, can be completely obscured by the shadow cast by the sidewalls of the grooves and the visual interference of the texture itself when they fall at the bottom of the grooves. Conventional visual inspection systems are limited by such blind spots and can hardly distinguish hidden weld lines from normal textures, resulting in parts that appear to have perfect surfaces but actually harbor hidden strength defects.

[0003] While existing injection molding machines can detect abnormalities at the melt junction using pressure or temperature sensors, indirectly providing a warning of potential weld defects, this indirect signal cannot provide direct evidence of the existence, location, and severity of defects. This creates a dilemma for on-site handling: ignoring the warning risks releasing parts with strength issues to the vehicle assembly stage; blindly adjusting parameters such as mold temperature and injection speed based solely on indirect signals may damage the originally acceptable texture appearance, leading to new batch appearance defects even when no actual defects exist. How to penetrate the visual blind spot of the texture, identify hidden weld defects, and implement targeted process compensation accordingly has become a pressing problem in this field.

[0004] To address this, the present invention provides a method for visual inspection of defects in injection-molded interior parts and adaptive control of process parameters. Summary of the Invention

[0005] The purpose of this invention is to provide a method for visual inspection of defects in injection-molded interior parts and adaptive control of process parameters to solve the aforementioned background problems.

[0006] The objective of this invention can be achieved through the following technical solution: a method for visual inspection of defects in injection-molded interior parts and adaptive control of process parameters, comprising:

[0007] Multi-point pressure response sequences at the weld line during the injection filling stage are collected, and the energy characteristics are separated by time period to obtain the attenuation metric vector. Based on the attenuation metric vector, deviation fusion analysis is performed to obtain the weld weakness index and generate the weld weakness index distribution.

[0008] Obtain grayscale images of the textured surface of the interior panel, and divide the visually shielded areas formed by valley occlusion according to the texture design drawings; spatially register the visually shielded areas with the distribution of weld weakness index to identify candidate areas of hidden defects;

[0009] Auxiliary lighting is applied to the candidate area of ​​hidden defects and a visible image is acquired. Strip feature enhancement processing is performed to obtain a gradient amplitude image. Candidate strip anomaly features are screened and geometric morphology descriptors are obtained. Morphological deviation analysis is performed based on the geometric morphology descriptors of the candidate strip anomaly features to determine whether there is a real weld defect.

[0010] If it exists, the attenuation degree of the attenuation metric vector corresponding to the actual weld defect is extracted, and the corresponding mold temperature control unit and injection speed curve segment are traced back. Temperature correction instructions and injection speed adjustment instructions are generated based on the pre-built response surface model.

[0011] Furthermore, the process of separating energy characteristics by time period is as follows:

[0012] Identify the impact fusion stage and the cooling relaxation stage based on the pressure response sequence;

[0013] In the impact fusion section, the pressure value of each sampling point in the pressure response sequence is subtracted from the cavity reference pressure to obtain the standard pressure response sequence. The square value of the pressure value corresponding to each sampling point in the standard pressure response sequence is calculated and integrated over time to obtain the impact fusion energy value.

[0014] During the cooling relaxation phase, the square value of the pressure value corresponding to each sampling point in the standard pressure response sequence is calculated, and the obtained square value sequence is fitted with a first-order exponential decay model using least squares to obtain the relaxation decay rate.

[0015] The impact fusion energy value and relaxation attenuation rate at the same measurement location are combined into a two-dimensional attenuation metric vector.

[0016] Furthermore, the process of identifying the impact fusion stage and the cooling relaxation stage is as follows:

[0017] The pressure change rate curve is obtained by taking the first derivative of the pressure response sequence, and the positive maximum point is searched. The time corresponding to the positive maximum point is marked as the start time of the impact fusion segment.

[0018] Starting from the beginning of the impact fusion segment, when the pressure value in the pressure response sequence is monitored to continue to decrease within a preset number of sampling points without any rebound fluctuation, the time corresponding to the sampling point that first meets the continuous decreasing condition is determined as the end time of the impact fusion segment.

[0019] The end time of the impact fusion section is taken as the start time of the cooling relaxation section, and the time when the pressure value continues to decrease to the cavity reference pressure set by the injection molding process is taken as the end time of the cooling relaxation section.

[0020] Furthermore, the process of generating the weld weakness index is as follows:

[0021] Based on the attenuation metric vector at each measurement location, the spatial interpolation method of radial basis function is used to obtain the attenuation metric vector at any spatial location on each weld line, and the ideal attenuation vector is set.

[0022] The attenuation metric vector is compared with the ideal attenuation vector to obtain the first deviation value and the second deviation value. The first deviation value and the second deviation value are combined to obtain the weld weakness index.

[0023] Furthermore, the process of delineating the visually shielded area formed by the valley is as follows:

[0024] Based on the texture design drawings, read the design direction and depth of the texture grooves, take the design direction as the trend of the texture grooves, and combine it with the spatial angle with the direction of parallel light source illumination, and use the geometric ray model to calculate the occlusion width.

[0025] Based on the internal and external parameters of the industrial camera, the masking width is projected onto the grayscale image to obtain the masking pixel band, and the center line of the texture groove is obtained and projected onto the grayscale image to obtain the center line pixel sequence. The union of the masking pixel band and the center line pixel sequence is used as the visual masking sub-region of the texture groove, and the union of all visual masking sub-regions is obtained to form the visual masking region.

[0026] Furthermore, the process of spatial registration is as follows:

[0027] The three-dimensional geometric data of the surface texture of the injection mold cavity are obtained and parametric unfolding is performed to generate a two-dimensional unfolded coordinate network;

[0028] The distribution of the weld weakness index is mapped to a two-dimensional unfolded coordinate network to form a weakness index map, and the visual shielding area is transformed to a two-dimensional unfolded coordinate network through the camera's internal and external parameters.

[0029] The nodes to be analyzed are extracted based on a two-dimensional unfolded coordinate network and the upper quartiles are determined by arranging them in ascending order. Weak and high nodes are identified based on the upper quartiles. The intersection of the continuous region formed by all weak and high nodes and the visual shielding region is obtained to obtain the candidate region of hidden defects.

[0030] Furthermore, the process of performing morphological deviation analysis is as follows:

[0031] After acquiring and displaying the image, strip feature enhancement processing is performed to select candidate strip anomaly features;

[0032] Extract the geometric morphology descriptors of each candidate strip-shaped anomaly feature, and extract the geometric morphology descriptors of regular textures in the texture design drawings. Compare the morphological deviations of the geometric morphology descriptors of the candidate strip-shaped anomaly features with those of the regular textures, and determine if they are strip-shaped anomalies. If the weld weakness index at the spatial location corresponding to the strip-shaped anomaly feature belongs to a weak high node and is located within the hidden defect candidate area, then the strip-shaped anomaly feature is determined to be a real weld defect.

[0033] Furthermore, the method for selecting candidate strip-shaped anomaly features is as follows:

[0034] Calculate the gradient direction and gradient magnitude of each pixel in the displayed image, collect the angle between the gradient direction of each pixel and the main extension direction of the texture groove, and identify pixels that follow the texture based on the angle.

[0035] The gradient magnitude of pixels with the same texture is preserved, while the gradient magnitude of pixels other than those with the same texture is set to zero, resulting in a gradient magnitude image after directional filtering.

[0036] For gradient magnitude images, a connected component analysis algorithm is used to extract connected components, and strip-shaped regions with a length exceeding the period length of a single texture groove are selected as candidate strip-shaped anomaly features.

[0037] Furthermore, the process of performing morphological deviation comparison is as follows:

[0038] Directional deviation comparison: obtain the directional deviation angle between the main extension direction of the candidate strip-shaped anomaly feature and the main extension direction of the texture groove; obtain the intersection angle between the predicted direction of the welded trace and the main extension direction of the texture groove; calculate the absolute difference between the directional deviation angle and the intersection angle; if the absolute difference is less than or equal to the angle deviation tolerance, the directional deviation condition is met.

[0039] Length deviation comparison: when a candidate strip-shaped anomaly feature continuously spans at least two texture grooves, the length deviation condition is met.

[0040] Width deviation comparison: Calculate the difference between the width of the candidate strip-shaped anomaly feature and the design width of the texture groove. If the difference exceeds the width tolerance range of the texture groove, the width deviation condition is met.

[0041] Furthermore, the process of generating temperature correction commands and injection speed adjustment commands is as follows:

[0042] Extract the attenuation degree of the attenuation metric vector corresponding to the actual weld defect as the expected compensation amount, and input it into the response surface model to solve in reverse to obtain the temperature change and injection speed change.

[0043] Temperature correction commands are generated for the mold temperature control unit based on temperature changes, and injection speed adjustment commands are generated for the injection molding machine control system based on injection speed changes.

[0044] The beneficial effects of this invention are as follows:

[0045] 1. Collect multi-point pressure response sequences at the weld line during the injection molding filling stage, perform time-segmented energy feature separation to obtain attenuation metric vectors, and perform deviation fusion analysis based on the attenuation metric vectors to obtain weld weakness indexes and generate weld weakness index distributions; transform the fleeting pressure fluctuations at melt convergence into a continuously distributed weakness index field along the entire weld line, giving previously invisible internal strength hazards a quantified spatial risk distribution map; acquire grayscale images of the interior panel texture surface, and divide the visual shielding areas formed by valley shading according to the texture design drawings; spatially register the visual shielding areas with the weld weakness index distribution to identify hidden defect candidate areas; through spatial registration of process data and visual information, break the visual deception caused by texture valley shadows, and delineate the truly high-risk points to be investigated from the general indirect warnings of sensors that are covered by the illusion of a perfect appearance.

[0046] 2. Apply auxiliary lighting to the candidate area of ​​hidden defects and acquire display images. Perform strip feature enhancement processing to obtain gradient amplitude images, screen candidate strip anomalies and obtain geometric morphology descriptors. Perform morphological deviation analysis based on the geometric morphology descriptors of the candidate strip anomalies to determine whether there are real weld defects. Expose hidden structures in the visual blind zone by actively changing the illumination. Combine morphological feature comparison to map ambiguous indirect risk signals into conclusive direct evidence of the existence of defects. If there are defects, extract the attenuation degree of the attenuation metric vector corresponding to the real weld defect and trace the corresponding mold temperature control unit and injection speed curve segment. Generate temperature correction instructions and injection speed adjustment instructions based on the pre-constructed response surface model. After confirming the real defect, quantitatively trace the local mold temperature and injection speed that caused the defect based on the attenuation degree and implement targeted and precise compensation. This will eliminate the hidden danger of weld strength while protecting the current qualified large-area texture appearance and prevent secondary defects in batch appearance caused by blind global adjustments. Attached Figure Description

[0047] The invention will now be further described with reference to the accompanying drawings.

[0048] Figure 1 This is a flowchart of the steps of the method for visual inspection of defects and adaptive control of process parameters in injection-molded interior parts according to the present invention;

[0049] Figure 2 This is a logic diagram for identifying strip-shaped abnormal features in this invention;

[0050] Figure 3This is a functional module diagram of the visual inspection and adaptive control system for defects in injection-molded interior parts in this invention. Detailed Implementation

[0051] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0052] Example 1

[0053] like Figure 1 As shown, the method for visual inspection of defects and adaptive adjustment of process parameters in injection-molded interior parts includes:

[0054] Step 1: Collect multi-point pressure response sequences at the weld line during the injection filling stage, perform time-segmented energy feature separation to obtain the attenuation metric vector, perform deviation fusion analysis based on the attenuation metric vector to obtain the weld weakness index and generate the weld weakness index distribution;

[0055] In the injection mold design stage, mold flow analysis is performed to indicate the predicted direction of each weld line; based on the predicted direction of each weld line, high temperature resistant miniature pressure sensors are installed at equal intervals, and the pressure sensors are used to collect real-time pressure signals at the weld lines during the injection filling stage to form a pressure response sequence at each measurement position.

[0056] Among them, the pressure response sequence fully records the entire process of pressure change from the melt front meeting to the completion of the merger and then to cooling and solidification;

[0057] For the pressure response sequence at each measurement location, time-segmented energy characteristic separation is automatically performed based on the physical stages of melt convergence behavior, specifically:

[0058] The pressure change rate curve is obtained by taking the first derivative of the pressure response sequence. The positive maximum point generated by the violent collision of the two melt fronts is searched on the pressure change rate curve. The time corresponding to the positive maximum point is marked as the start time of the impact fusion segment.

[0059] From the start of the impact fusion phase, the pressure value of the pressure response sequence rises sharply and forms a steep pulse, after which the pressure value continues to decrease. When the pressure value in the pressure response sequence is monitored to continue to decrease within a preset number of sampling points without any rebound fluctuation, the time corresponding to the sampling point that first meets this continuous decreasing condition is determined as the end time of the impact fusion phase.

[0060] The end time of the impact fusion segment is taken as the start time of the cooling relaxation segment, and the time when the pressure value in the pressure response sequence continues to decay to the cavity reference pressure set by the injection molding process is taken as the end time of the cooling relaxation segment. Thus, the pressure response sequence is divided into the impact fusion segment and the cooling relaxation segment.

[0061] It is understandable that the cavity reference pressure refers to the stable pressure reference value that should be returned to by the melt in the cavity after filling and cooling to an equilibrium state, which is preset in the injection molding process.

[0062] In the impact fusion section, the pressure value of each sampling point in the pressure response sequence is subtracted from the cavity reference pressure to obtain the standard pressure response sequence. The square value of the pressure value corresponding to each sampling point in the standard pressure response sequence is calculated and integrated over time to obtain the impact fusion energy value.

[0063] During the cooling relaxation phase, the square value of the pressure value corresponding to each sampling point in the standard pressure response sequence is calculated, and the obtained square value sequence is fitted with a first-order exponential decay model using least squares to obtain the relaxation decay rate.

[0064] It should be noted that, according to the formation mechanism of injection molding weld lines, the larger the impact fusion energy value, the higher the temperature and momentum when the melts meet, and the more fully the oxide layer is broken and the interface is tightly contacted; the smaller the relaxation decay rate, the longer the high temperature is maintained after the melts merge, and the more relaxed the molecular chain diffusion and entanglement. Both of these factors together determine the bonding strength of the weld interface.

[0065] By combining the impact fusion energy value and relaxation attenuation rate at the same measurement location, a two-dimensional attenuation metric vector is constructed. The attenuation metric vector quantitatively expresses the potential attenuation degree of the bonding strength of the fusion interface at the measurement location relative to the ideal fusion state.

[0066] The spatial interpolation method of radial basis function is adopted to continuously extend and map the attenuation metric vectors of each discrete measurement position along their respective weld lines, so that any spatial position on each weld line has an attenuation metric vector obtained by spatial interpolation.

[0067] Based on the pressure-volume-temperature characteristics of the polymer used in injection molding, the specific volume change of the melt is determined. According to the molecular diffusion coefficient and the weld healing model, the theoretical impact fusion energy value and theoretical relaxation attenuation rate that the weld trace at any spatial location should have when fully ideally fused are calculated under the ideal mold temperature field and optimal melt temperature conditions. Thus, the ideal attenuation vector is constructed.

[0068] For the attenuation measurement vector at each spatial position on each weld line, compare it with the ideal attenuation vector at the corresponding spatial position: calculate the degree of deviation between the impact fusion energy value and the theoretical impact fusion energy value, and the degree of deviation between the relaxation attenuation rate and the theoretical relaxation attenuation rate, to obtain the first deviation value and the second deviation value.

[0069] The first deviation value and the second deviation value are combined to obtain the weld weakness index at that spatial location of the weld trace;

[0070] The process of combining the values ​​is as follows: the first deviation value and the second deviation value are normalized according to their ranges so that they are both mapped to the interval [0,1]; the two normalized values ​​are multiplied directly, and the product is used as the welding weakness index at that spatial location.

[0071] The weld weakness index of all spatial locations on all weld lines is uniformly expressed in the three-dimensional numerical field under the cavity coordinate system, generating the weld weakness index distribution of each weld line at the cavity scale.

[0072] Step 2: Obtain a grayscale image of the textured surface of the interior panel, and divide the visual shielding area formed by the valley occlusion according to the texture design drawings; perform spatial position registration between the visual shielding area and the weld weakness index distribution to identify candidate areas of hidden defects;

[0073] A grayscale image of the textured surface of the interior panel after molding is acquired. This grayscale image is taken by an industrial camera installed at the standard inspection station in the injection molding workshop. During the shooting, a parallel light source with a known fixed angle to the main extension direction of the texture groove is used to provide uniform illumination, so that the bottom of the texture groove forms a stable shadow due to directional occlusion.

[0074] Obtain the texture design drawing on which the texture surface of the injection mold cavity is based during the processing stage; read the design direction and design depth of the texture groove based on the texture design drawing; and take the design direction as the trend of the texture groove.

[0075] Based on the trend of the texture grooves, the visual shielding area formed by the valley occlusion is divided: for each texture groove read from the texture design drawing, according to its design depth, the spatial angle between the illumination direction of the parallel light source and the trend of the texture groove, the occlusion width projected by one side wall of the groove at the bottom of the valley is calculated using a geometric ray model based on parallel light occlusion.

[0076] It should be noted that the geometric ray model is based on the fundamental principle of light propagating in straight lines. In the geometric ray model, a parallel light source illuminates the surface of the interior panel at a preset fixed angle, and the light rays intersect spatially with the sidewalls of the texture grooves. Due to the shading of the groove sidewalls, some incident light rays cannot reach the bottom of the groove, thus forming a continuous shadow area at the bottom of the groove. The width of this shadow area along the direction perpendicular to the texture direction is uniquely determined by the spatial relative relationship between the incident direction of the light rays and the geometric shape of the texture groove: the larger the angle between the incident light rays and the normal of the cavity surface, or the greater the groove depth, the wider the shading width; conversely, the narrower the shading width. Using the geometric ray model, based on the known illumination angle and texture design parameters, the shading range that each texture groove will inevitably exist in the detection image, i.e., the visual shielding area, is determined in advance.

[0077] Obtain the intrinsic and extrinsic parameters of the industrial camera, project the occlusion width onto the pixel coordinate system of the grayscale image, obtain the continuous set of pixels corresponding to the occlusion width in the grayscale image, and define it as the occlusion pixel band.

[0078] The center line of the texture groove in the texture design drawing is projected onto the pixel coordinate system of the grayscale image according to the same intrinsic and extrinsic parameters of the industrial camera, so as to obtain the continuous set of pixels corresponding to the center line of the texture groove in the grayscale image, and defined as the center line pixel sequence.

[0079] All pixels corresponding to the union of the masking pixel band and the center line pixel sequence are defined as the visual masking sub-region of the texture groove in the grayscale image.

[0080] Take the union of all visually shielded sub-regions corresponding to the texture grooves to generate the complete visually shielded region;

[0081] The weak weld index distribution of each weld line obtained in step one is spatially registered with the visual shielding area, specifically as follows:

[0082] Obtain a three-dimensional digital model of the injection mold cavity, and extract three-dimensional geometric data of the textured surface of the injection mold cavity based on the three-dimensional digital model. The three-dimensional geometric data is the three-dimensional coordinates of the vertices of all triangular facets that constitute the textured surface of the injection mold cavity.

[0083] Parametric unfolding is performed on the three-dimensional geometric data of the injection mold cavity texture surface: the three-dimensional geometric data of the injection mold cavity texture surface is taken as input, and a mesh parametric algorithm based on least squares conformal mapping is adopted to maintain the angular relationship between each triangular facet on the injection mold cavity texture surface and minimize the side length deformation. The three-dimensional geometric data of the injection mold cavity texture surface is flattened onto a two-dimensional plane to generate a two-dimensional unfolded coordinate network that corresponds one-to-one with the injection mold cavity texture surface geometrically.

[0084] Each node of the two-dimensional unfolded coordinate network uniquely corresponds to a spatial position on the surface of the injection mold cavity texture, thus completely mapping any point on the surface of the injection mold cavity texture to the two-dimensional unfolded coordinate network.

[0085] The weld weakness index distribution in the cavity coordinate system generated in step one is projected onto the two-dimensional unfolded coordinate network according to the mapping relationship between the three-dimensional geometric data and the two-dimensional unfolded coordinate network, and the weakness index map covering the textured surface of the interior panel after molding is obtained by centroid interpolation.

[0086] By using the intrinsic and extrinsic parameters of the industrial camera, a corresponding mapping relationship is established between each pixel of the grayscale image and each node in the two-dimensional unfolded coordinate network. Based on the corresponding mapping relationship, the visual shielding area of ​​the grayscale image is transformed into the two-dimensional unfolded coordinate network, so that the visual shielding area and the weakness index map are in the same space.

[0087] Extract all nodes with dual attributes from the two-dimensional unfolded coordinate network and mark them as nodes to be analyzed;

[0088] Among them, the first attribute of the dual attributes is that the node is in a visually shielded area, and the second attribute is that the node has a weld weakness index.

[0089] Obtain the weld weakness index of all nodes to be analyzed, sort them in ascending order, and determine the upper quartile; identify nodes with a weld weakness index greater than the upper quartile as nodes with relatively high weakness.

[0090] The intersection of the continuous region formed by all weak and high nodes with the visual shielding region is obtained, and each connected region is the candidate region for hidden defects.

[0091] The technical solution of this invention is as follows: A multi-point pressure response sequence at the weld line during the injection molding filling stage is collected; time-segmented energy feature separation is performed to obtain an attenuation metric vector; deviation fusion analysis is performed based on the attenuation metric vector to obtain a weld weakness index and generate a weld weakness index distribution; the fleeting pressure fluctuations during melt fusion are transformed into a continuously distributed weakness index field along the entire weld line, giving previously invisible internal strength hazards a quantified spatial risk distribution map; a grayscale image of the interior panel texture surface is acquired, and visual shielding areas formed by valley shading are delineated based on the texture design drawings; the spatial registration of the visual shielding areas with the weld weakness index distribution identifies hidden defect candidate areas; through spatial registration of process data and visual information, the visual deception caused by texture valley shadows is broken, and the truly high-risk inspection points concealed by the illusion of a perfect appearance are identified from the general indirect warnings of sensors.

[0092] Example 2

[0093] Please see Figures 1-2 As shown, the method for visual inspection of defects and adaptive adjustment of process parameters in injection-molded interior parts includes:

[0094] Step 3: Apply auxiliary lighting to the candidate area of ​​hidden defects and acquire the image of the appearance. Perform strip feature enhancement processing to obtain the gradient amplitude image, screen candidate strip abnormal features and obtain geometric shape descriptors. Perform morphological deviation analysis based on the geometric shape descriptors of the candidate strip abnormal features to determine whether there is a real welding defect.

[0095] Auxiliary illumination is applied to the candidate region of hidden defects, and revealing images are acquired, specifically as follows:

[0096] Switch the parallel light source at the standard inspection station in the injection molding workshop to a non-parallel light source, so that the angle between the illumination projection direction of the non-parallel light source and the main extension direction of the texture grooves on the interior panel surface is not equal to the known fixed angle between the parallel light source and the main extension direction of the texture grooves.

[0097] Among them, the angle between the illumination projection direction of the non-parallel light source and the main extension direction of the texture groove is set so that the light emitted by the non-parallel light source can directly enter the bottom of the groove of the texture groove, and no longer form a shielding width at the bottom of the groove due to the obstruction of the sidewall of the texture groove.

[0098] Under non-parallel light source illumination conditions, an industrial camera is used to photograph each hidden defect candidate area one by one to obtain a visible image of the hidden defect candidate area.

[0099] It should be noted that, because light from non-parallel light sources can reach the bottom of the valley, the surface morphology of the valley bottom, which is completely covered by shadows under parallel light sources, can be presented in the display image; when there is a weld line at the bottom of the valley in the candidate area of ​​the hidden defect, the surface micro-undulations formed by the polymer material at the weld line interface with the melt convergence interface create a reflection difference with the surrounding normal texture surface. This reflection difference manifests as a local grayscale abnormality in the display image under non-parallel light source illumination.

[0100] For each displayed image, perform strip feature enhancement processing based on the consistency between the image gradient direction and the main extension direction of the texture grooves;

[0101] The specific process of the strip feature enhancement processing is as follows: calculate the gradient direction and gradient magnitude of each pixel in the displayed image, collect the angle between the gradient direction of each pixel and the main extension direction of the texture groove, and if the angle is less than or equal to the allowable value of the direction deviation, then the corresponding pixel is identified as a pixel with the same texture.

[0102] It should be noted that the allowable value of directional deviation is determined based on the statistical distribution of the pixel-level direction of the texture groove design in the grayscale image. All pixels located on the pixel sequence of the texture groove center line in the grayscale image are extracted, and the angle between the gradient direction of each pixel and the main extension direction of the texture groove is calculated. The standard deviation of all the obtained angles is used as the allowable value of directional deviation.

[0103] The gradient magnitude of pixels with the same texture is preserved, while the gradient magnitude of pixels other than those with the same texture is set to zero, resulting in a gradient magnitude image after directional filtering.

[0104] It should be noted that in the gradient magnitude image, the edge of the texture groove itself is completely preserved because its gradient direction is consistent with the main extension direction of the texture groove, while the weld line appears as a strip structure at a certain angle to the main extension direction of the texture groove, and there is a stable directional difference between its gradient direction and the main extension direction of the texture groove.

[0105] For gradient magnitude images, a connected component analysis algorithm is used to extract connected components, and strip-shaped regions with a length exceeding the period of a single texture groove are selected as candidate strip-shaped anomaly features.

[0106] The period length of a single texture groove is read from the texture design drawing and is equal to the distance between the center lines of two adjacent texture grooves.

[0107] Extract the geometric morphological descriptor of each candidate strip anomaly feature, including the main extension direction of the candidate strip anomaly feature, the length of the candidate strip anomaly feature, and the width of the candidate strip anomaly feature;

[0108] Using the regular texture defined by the texture design drawing within the hidden defect candidate area as a reference, the geometric shape descriptor of the regular texture is extracted from the texture design drawing. The geometric shape descriptor of the regular texture includes the main extension direction of the texture groove, the period length of the texture groove, and the design width of the texture groove.

[0109] The geometric morphological descriptors of candidate strip-shaped anomalies are compared with the geometric morphological descriptors of regular textures to determine morphological deviations.

[0110] Among them, the shape deviation comparison includes direction deviation comparison, length deviation comparison and width deviation comparison;

[0111] Directional deviation comparison: Calculate the directional deviation angle between the main extension direction of the candidate strip-shaped anomaly features and the main extension direction of the texture groove; calculate the intersection angle between the predicted direction of each weld line and the main extension direction of the texture groove.

[0112] The absolute difference between the directional deviation angle and the intersection angle is calculated. If the result of the absolute difference calculation is less than or equal to the angle deviation tolerance, the candidate strip-shaped anomaly feature is determined to meet the directional deviation condition.

[0113] The angular deviation tolerance is determined based on the fluctuation statistical range of each weld line direction in the mold flow analysis;

[0114] Length deviation comparison: When a candidate strip-shaped anomaly feature continuously crosses two or more texture grooves, the candidate strip-shaped anomaly feature is determined to meet the length deviation condition.

[0115] Width Deviation Comparison: Calculate the difference between the width of the candidate strip-shaped anomaly feature and the design width of the texture groove. When the difference exceeds the width tolerance range of the texture groove in the texture design drawing, the candidate strip-shaped anomaly feature is determined to meet the width deviation condition.

[0116] When a candidate strip-shaped anomaly simultaneously satisfies the direction deviation condition, length deviation condition, and width deviation condition, the candidate strip-shaped anomaly is determined to be a strip-shaped anomaly reflecting the weld line morphology.

[0117] Extract the pixel coordinates of the strip-shaped anomaly features in the displayed image. Based on the intrinsic and extrinsic parameters of the industrial camera and the corresponding mapping relationship between the two-dimensional unfolded coordinate network and each pixel of the grayscale image, back-project the pixel coordinates of the strip-shaped anomaly features in the displayed image to the cavity coordinate system to obtain the spatial position of the strip-shaped anomaly features.

[0118] Compare the spatial location corresponding to the strip-shaped anomaly with the distribution of the weld weakness index. If the weld weakness index at the spatial location corresponding to the strip-shaped anomaly belongs to the weak high node and is located in the hidden defect candidate area, then the strip-shaped anomaly is determined to be a real weld defect, and a real weld defect exists in the hidden defect candidate area.

[0119] When a candidate strip-shaped anomaly does not meet any of the conditions for directional deviation, length deviation, or width deviation, the candidate strip-shaped anomaly is determined to be a local morphological fluctuation of the texture groove itself, and there is no real welding defect in the candidate area of ​​this hidden defect.

[0120] Step 4: If it exists, extract the attenuation degree of the attenuation metric vector corresponding to the actual weld defect, trace the corresponding mold temperature control unit and injection speed curve segment, and generate temperature correction instructions and injection speed adjustment instructions based on the pre-built response surface model.

[0121] If it is determined that there is a real weld defect in the hidden defect candidate area, the attenuation metric vector corresponding to the real weld defect is read from the attenuation metric vector generated in step one, according to the spatial position of the real weld defect obtained in step three, and its corresponding ideal attenuation vector is obtained.

[0122] Calculate the first and second deviation values ​​between the attenuation metric vector corresponding to the actual weld defect and the ideal attenuation vector, and use the first and second deviation values ​​together as the attenuation degree of the attenuation metric vector;

[0123] The mold temperature control unit is pre-built in the three-dimensional digital model of the injection mold cavity. The mold temperature control unit is a heating or cooling unit set inside the injection mold, each with an independent temperature control loop and used to adjust the local surface temperature of the mold cavity. The mold temperature control unit has a one-to-one correspondence with the weld line.

[0124] Match the spatial location of the actual weld defect with the predicted direction of each weld trace in step one to determine the weld trace to which the spatial location of the actual weld defect belongs, and read the mold temperature control unit corresponding to the actual weld defect based on the correspondence between the mold temperature control unit and the weld trace.

[0125] The injection speed curves recorded in the injection molding machine control system are obtained, which are actually executed during the current injection cycle. The injection speed curves divide the entire injection process into several continuous injection segments with time as the horizontal axis. Each injection segment has a segment identifier and a preset injection speed value.

[0126] Obtain the start time of the impact fusion segment, determine it as the physical moment when the leading edges of the two melts meet at the spatial position, and mark it as the meeting moment;

[0127] Mark the meeting point on the time axis of the multi-segment injection speed curve, read the injection segment where the meeting point falls, and mark it as the fusion-dominant injection segment.

[0128] The injection speed value and segment identification of the main injection segment of the fusion are obtained based on the multi-segment injection speed curve;

[0129] The attenuation degree (first deviation value and second deviation value) of the attenuation metric vector of the actual weld defect is used as the expected compensation amount. It is then input into the response surface model for inverse solution to calculate the temperature change of the mold temperature control unit and the injection speed change of the injection speed curve segment required to bring the first deviation value and the second deviation value to zero.

[0130] It should be noted that the pre-establishment process of the response surface model is as follows: multiple sets of process experiments are conducted using the same injection mold and material system as the current production. In each set of process experiments, the set temperature of the mold temperature control unit, which plays a dominant role, and the injection speed value of the injection speed curve segment are changed independently by a fixed amount. The impact fusion energy value and relaxation attenuation rate at the corresponding spatial location of the actual weld defect under each set of process experiment conditions are obtained using the methods in steps one and two, and the first deviation value and the second deviation value are calculated. The change in the set temperature of the mold temperature control unit and the change in the injection speed value of the injection speed curve segment are used as inputs, and the changes in the first deviation value and the second deviation value are used as outputs. A response surface model reflecting the quantitative relationship between the mold temperature and injection speed and the change in the bonding strength of the weld interface is constructed through multivariate nonlinear regression.

[0131] Based on the calculated temperature change of the mold temperature control unit, a temperature correction instruction is generated for the mold temperature control unit corresponding to the actual welding defect. The temperature correction instruction includes the identifier of the mold temperature control unit and the temperature change value.

[0132] Based on the calculated change in injection speed, an injection speed adjustment command is generated for the main injection segment corresponding to the actual weld defect.

[0133] The temperature correction command is sent to the temperature controller connected to the mold temperature control unit, and the injection speed adjustment command is sent to the injection molding machine control system, and implemented at the beginning of the next injection cycle;

[0134] The technical solution of this invention is as follows: Auxiliary lighting is applied to the candidate area of ​​hidden defects, and a visible image is acquired. Strip feature enhancement processing is performed to obtain a gradient amplitude image. Candidate strip-shaped abnormal features are screened, and geometric morphological descriptors are obtained. Morphological deviation analysis is performed based on the geometric morphological descriptors of the candidate strip-shaped abnormal features to determine whether a real weld defect exists. Hidden structures in the visual blind zone are exposed by actively changing the illumination. Combined with morphological feature comparison, ambiguous indirect risk signals are mapped into conclusive direct evidence of the existence of defects. If a defect exists, the attenuation degree of the attenuation metric vector corresponding to the real weld defect is extracted, and the corresponding mold temperature control unit and injection speed curve segment are traced. Temperature correction instructions and injection speed adjustment instructions are generated based on a pre-constructed response surface model. After confirming the real defect, the local mold temperature and injection speed causing the defect are quantitatively traced based on the attenuation degree, and targeted and precise compensation is implemented. This eliminates potential weld strength hazards while perfectly protecting the current qualified large-area texture appearance, preventing secondary defects in batch appearance caused by blind global adjustments.

[0135] Example 3

[0136] Please see Figure 3 As shown, the visual inspection system for defects in injection-molded interior parts and the adaptive control system for process parameters include the following modules:

[0137] Melt mark assessment module: Collects multi-point pressure response sequences at the weld line during the injection filling stage, performs time-segmented energy feature separation to obtain attenuation metric vector, performs deviation fusion analysis based on attenuation metric vector to obtain weld weakness index and generate weld weakness index distribution;

[0138] Concealed defect identification module: acquires grayscale images of the textured surface of the interior panel, and divides the visually shielded area formed by the valley occlusion according to the texture design drawings; spatially registers the visually shielded area with the distribution of the weld weakness index to identify candidate areas of concealed defects;

[0139] The display and discrimination module applies auxiliary lighting to the candidate area of ​​hidden defects and acquires the display image. It performs strip feature enhancement processing to obtain the gradient amplitude image, filters candidate strip abnormal features and obtains geometric shape descriptors. Based on the geometric shape descriptors of the candidate strip abnormal features, it performs morphological deviation analysis to determine whether there is a real welding defect.

[0140] Targeted compensation module: If it exists, it extracts the attenuation degree of the attenuation metric vector corresponding to the actual weld defect, traces the corresponding mold temperature control unit and injection speed curve segment, and generates temperature correction instructions and injection speed adjustment instructions based on the pre-built response surface model.

[0141] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the present invention should still fall within the scope of the present invention.

Claims

1. A method for visual inspection of defects and adaptive adjustment of process parameters in injection-molded interior parts, characterized by: include: Multi-point pressure response sequences at the weld line during the injection filling stage are collected, and the energy characteristics are separated by time period to obtain the attenuation metric vector. Based on the attenuation metric vector, deviation fusion analysis is performed to obtain the weld weakness index and generate the weld weakness index distribution. Obtain grayscale images of the textured surface of the interior panel, and divide the visually shielded areas formed by valley occlusion according to the texture design drawings; spatially register the visually shielded areas with the distribution of weld weakness index to identify candidate areas of hidden defects; Auxiliary lighting is applied to the candidate area of ​​hidden defects and a visible image is acquired. Strip feature enhancement processing is performed to obtain a gradient amplitude image. Candidate strip anomaly features are screened and geometric morphology descriptors are obtained. Morphological deviation analysis is performed based on the geometric morphology descriptors of the candidate strip anomaly features to determine whether there is a real weld defect. If it exists, the attenuation degree of the attenuation metric vector corresponding to the actual weld defect is extracted, and the corresponding mold temperature control unit and injection speed curve segment are traced back. Temperature correction instructions and injection speed adjustment instructions are generated based on the pre-built response surface model.

2. The method for visual inspection of defects and adaptive control of process parameters in injection-molded interior parts according to claim 1, characterized in that: The process of performing time-segmented energy feature separation is as follows: Identify the impact fusion stage and the cooling relaxation stage based on the pressure response sequence; In the impact fusion section, the pressure value of each sampling point in the pressure response sequence is subtracted from the cavity reference pressure to obtain the standard pressure response sequence. The square value of the pressure value corresponding to each sampling point in the standard pressure response sequence is calculated and integrated over time to obtain the impact fusion energy value. During the cooling relaxation phase, the square value of the pressure value corresponding to each sampling point in the standard pressure response sequence is calculated, and the obtained square value sequence is fitted with a first-order exponential decay model using least squares to obtain the relaxation decay rate. The impact fusion energy value and relaxation attenuation rate at the same measurement location are combined into a two-dimensional attenuation metric vector.

3. The method for visual inspection of defects and adaptive adjustment of process parameters in injection-molded interior parts according to claim 2, characterized in that: The process of identifying the impact fusion section and the cooling relaxation section is as follows: The pressure change rate curve is obtained by taking the first derivative of the pressure response sequence, and the positive maximum point is searched. The time corresponding to the positive maximum point is marked as the start time of the impact fusion segment. Starting from the beginning of the impact fusion segment, when the pressure value in the pressure response sequence is monitored to continue to decrease within a preset number of sampling points without any rebound fluctuation, the time corresponding to the sampling point that first meets the continuous decreasing condition is determined as the end time of the impact fusion segment. The end time of the impact fusion section is taken as the start time of the cooling relaxation section, and the time when the pressure value continues to decrease to the cavity reference pressure set by the injection molding process is taken as the end time of the cooling relaxation section.

4. The method for visual inspection of defects and adaptive control of process parameters in injection-molded interior parts according to claim 1, characterized in that: The process of generating the weld weakness index is as follows: Based on the attenuation metric vector at each measurement location, the spatial interpolation method of radial basis function is used to obtain the attenuation metric vector at any spatial location on each weld line, and the ideal attenuation vector is set. The attenuation metric vector is compared with the ideal attenuation vector to obtain the first deviation value and the second deviation value. The first deviation value and the second deviation value are combined to obtain the weld weakness index.

5. The method for visual inspection of defects and adaptive control of process parameters in injection-molded interior parts according to claim 1, characterized in that: The process of defining the visually shielded area formed by the valley is as follows: Based on the texture design drawings, read the design direction and depth of the texture grooves, take the design direction as the trend of the texture grooves, and combine it with the spatial angle with the direction of parallel light source illumination, and use the geometric ray model to calculate the occlusion width. Based on the internal and external parameters of the industrial camera, the masking width is projected onto the grayscale image to obtain the masking pixel band, and the center line of the texture groove is obtained and projected onto the grayscale image to obtain the center line pixel sequence. The union of the masking pixel band and the center line pixel sequence is used as the visual masking sub-region of the texture groove, and the union of all visual masking sub-regions is obtained to form the visual masking region.

6. The method for visual inspection of defects and adaptive control of process parameters in injection-molded interior parts according to claim 1, characterized in that: The process of spatial registration is as follows: The three-dimensional geometric data of the surface texture of the injection mold cavity are obtained and parametric unfolding is performed to generate a two-dimensional unfolded coordinate network; The distribution of the weld weakness index is mapped to a two-dimensional unfolded coordinate network to form a weakness index map, and the visual shielding area is transformed to a two-dimensional unfolded coordinate network through the camera's internal and external parameters. The nodes to be analyzed are extracted based on a two-dimensional unfolded coordinate network and the upper quartiles are determined by arranging them in ascending order. Weak and high nodes are identified based on the upper quartiles. The intersection of the continuous region formed by all weak and high nodes and the visual shielding region is obtained to obtain the candidate region of hidden defects.

7. The method for visual inspection of defects and adaptive control of process parameters in injection-molded interior parts according to claim 1, characterized in that: The process of performing morphological deviation analysis is as follows: After acquiring and displaying the image, strip feature enhancement processing is performed to select candidate strip anomaly features; Extract the geometric morphology descriptors of each candidate strip-shaped anomaly feature, and extract the geometric morphology descriptors of regular textures in the texture design drawings. Compare the morphological deviations of the geometric morphology descriptors of the candidate strip-shaped anomaly features with those of the regular textures, and determine if they are strip-shaped anomalies. If the weld weakness index at the spatial location corresponding to the strip-shaped anomaly feature belongs to a weak high node and is located within the hidden defect candidate area, then the strip-shaped anomaly feature is determined to be a real weld defect.

8. The method for visual inspection of defects and adaptive control of process parameters in injection-molded interior parts according to claim 7, characterized in that: The method for obtaining candidate bar-shaped anomaly features through screening is as follows: Calculate the gradient direction and gradient magnitude of each pixel in the displayed image, collect the angle between the gradient direction of each pixel and the main extension direction of the texture groove, and identify pixels that follow the texture based on the angle. The gradient magnitude of pixels with the same texture is preserved, while the gradient magnitude of pixels other than those with the same texture is set to zero, resulting in a gradient magnitude image after directional filtering. For gradient magnitude images, a connected component analysis algorithm is used to extract connected components, and strip-shaped regions with a length exceeding the period length of a single texture groove are selected as candidate strip-shaped anomaly features.

9. The method for visual inspection of defects and adaptive control of process parameters in injection-molded interior parts according to claim 7, characterized in that: The process of performing morphological deviation comparison is as follows: Directional deviation comparison: obtain the directional deviation angle between the main extension direction of the candidate strip-shaped anomaly feature and the main extension direction of the texture groove; obtain the intersection angle between the predicted direction of the welded trace and the main extension direction of the texture groove; calculate the absolute difference between the directional deviation angle and the intersection angle; if the absolute difference is less than or equal to the angle deviation tolerance, the directional deviation condition is met. Length deviation comparison: when a candidate strip-shaped anomaly feature continuously spans at least two texture grooves, the length deviation condition is met. Width deviation comparison: Calculate the difference between the width of the candidate strip-shaped anomaly feature and the design width of the texture groove. If the difference exceeds the width tolerance range of the texture groove, the width deviation condition is met.

10. The method for visual inspection of defects and adaptive control of process parameters in injection-molded interior parts according to claim 1, characterized in that: The process of generating temperature correction commands and injection speed adjustment commands is as follows: Extract the attenuation degree of the attenuation metric vector corresponding to the actual weld defect as the expected compensation amount, and input it into the response surface model to solve in reverse to obtain the temperature change and injection speed change. Temperature correction commands are generated for the mold temperature control unit based on temperature changes, and injection speed adjustment commands are generated for the injection molding machine control system based on injection speed changes.

Images