A cylinder head flatness detection device

Abstract

This invention discloses a cylinder head flatness detection device, comprising: a residual gradient construction module for acquiring the residuals of measuring points, dividing the grid, and calculating the mean of the grid residuals; a steep change region identification module for filtering steep change points and clustering them into steep change regions based on the gradient modulus, and extracting the region area and gradient direction angle; a direction sensitivity determination module for generating a direction sensitivity type based on the median of the gradient direction angle and the cosine of the angle between the assembly direction and the grid; a spatial autocorrelation analysis module for calculating the global mean and standard deviation based on the mean of the grid residuals, identifying clustered grids using the Moran's index, and determining the significance level; and a comprehensive level output module for multiplying the area, direction sensitivity weight, and cluster significance weight to obtain a level value, and comparing the results with a threshold to output the level. This invention integrates steep change identification, direction sensitivity, and spatial autocorrelation, employing multi-dimensional weighted grading to improve the detection of complex deformations and the ability to quantify risks.

Description

Technical Field

[0001] This invention relates to the field of testing equipment technology, and specifically to a cylinder head flatness testing device. Background Technology

[0002] In engine manufacturing, the flatness of the mating surface between the cylinder head and cylinder block is a key quality indicator for ensuring combustion chamber sealing and preventing leakage of high-temperature, high-pressure gases and coolant. Currently, cylinder head flatness is generally inspected using a coordinate measuring machine or non-contact optical scanning equipment to acquire surface point cloud data. Then, the least squares method is used to fit an ideal reference plane. The maximum deviation or total peak-to-valley value of all measuring points relative to this reference plane is taken as the flatness error, and it is compared with the design tolerance zone to determine whether the workpiece is qualified.

[0003] In existing technologies, cylinder head flatness inspection typically involves collecting height data from measuring points and calculating extreme deviations or root mean squares as the basis for judgment. This operational mode does not analyze the spatial distribution relationship and deformation direction characteristics between measuring points. In actual production, some cylinder heads may have overall deviations within the allowable range, but localized continuous undulations along the assembly force direction may occur. These directional hidden dangers are often missed because they are not reflected by the extreme deviation values, potentially leading to leaks during subsequent assembly. Furthermore, micron-level gradual warping caused by uneven casting cooling may have single-point deviations below the noise threshold, but adjacent measuring points exhibit a systematic positive trend. Existing independent threshold comparison methods cannot identify such clustered weak deformations. Moreover, the inspection results are only binary conclusions of pass or fail. Large-area, shallow-depth deformations and small-area, sharply deep dents may have similar overall deviations, but their impact on sealing performance differs. Factories lack a grading basis and can only handle them uniformly, increasing rework costs or quality risks. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and to propose a cylinder head flatness detection device.

[0005] The technical solution provided by this invention to solve the above problems is: a cylinder head flatness detection device, comprising a flatness detection system, wherein the detection system includes: The residual gradient construction module is used to obtain the residual value of the measuring point, divide it into a 10mm×10mm grid and calculate the mean value of the grid residual. For each measuring point, the eight-neighbor residual difference operation is performed to obtain an eight-dimensional vector as the measuring point gradient vector. The mean value of the grid residual is passed to the spatial autocorrelation analysis module and the measuring point gradient vector is passed to the steep change region identification module. The abrupt change region identification module is used to extract the magnitude based on the gradient vector of the measurement point and compare it with the threshold to filter abrupt change points. It uses the nearest neighbor clustering algorithm to cluster abrupt change region, calculates the region area and extracts the gradient direction angle, and passes the region area to the comprehensive level output module and the gradient direction angle to the direction sensitivity determination module. The orientation sensitivity determination module is used to obtain the main deformation direction based on the median of the gradient orientation angle in ascending order, calculate the cosine value of the angle between the orientation and the assembly orientation vector and compare it with 0.8 to generate the orientation sensitivity type, and then pass the orientation sensitivity type to the comprehensive level output module. The spatial autocorrelation analysis module is used to calculate the global mean and standard deviation based on the mean of grid residuals, use Moran's index to obtain statistics to compare critical values ​​and filter clustered grids, merge and remove small blocks, calculate the difference between the internal residual mean and the global mean, compare it with the standard deviation multiple to obtain clustered blocks and their significance levels, and pass the clustered blocks and their significance levels to the comprehensive level output module.

[0006] Preferably, it also includes a comprehensive grade output module, which receives the area, orientation sensitivity type, and clusters and their salience levels, generates defect blocks by spatial overlap matching, takes a coefficient of 1.2 or 0.5 based on the orientation sensitivity type, and takes a coefficient of 1.5, 1.0 or 0.5 based on the clusters and their salience levels, multiplies the area by the two coefficients to get the grade value, and compares it with the thresholds A and B to obtain the flatness detection grade, i.e., grade C is unqualified, grade B requires attention, or grade A is qualified, which is used as the system output result.

[0007] Preferably, the mean value of the grid residual includes local warping parameters and unit surface undulation, and the gradient vector of the measurement point includes microscopic physical potential and microstructural skewness.

[0008] Preferably, the area includes the physical sweep range and the geometric coverage scale, and the gradient direction angle includes the tilt azimuth parameter and the shape deflection attribute.

[0009] Preferably, the directional sensitivity type includes structural interlocking influence category and mechanical docking obstacle qualitative, and the aggregate and its significance level include contiguous mutation groups and local deterioration severity assessment conclusion.

[0010] Preferably, the residual gradient construction module includes: The numerical acquisition submodule obtains the residual values ​​of the measuring points, extracts the residual values ​​from the measurement data for each measuring point, and associates these values ​​with the plane coordinates of the measuring point to obtain the residual values ​​of the measuring points. The block mean submodule divides the measuring point plane into square grids with a side length of 10 mm based on the residual values ​​of the measuring points. It sums up the residual values ​​of all measuring points in each grid and divides them by the number of measuring points in that grid to obtain the grid mean residual value. The vector construction submodule, for each measurement point, calls the measurement point residual value and extracts the measurement point residual values ​​at eight positions: up, down, left, right, and four diagonal directions. Subtracting the measurement point residual value from each neighborhood residual value yields eight difference values. Arrange the eight difference values ​​into a vector in clockwise order starting from due north to obtain the measurement point gradient vector.

[0011] Preferably, the abrupt change region identification module includes: The steep change point filtering submodule calculates the Euclidean magnitude of each vector based on the gradient vector of the measurement point, compares each magnitude with a set magnitude threshold, and assigns steep change markers to measurement points whose magnitudes exceed the threshold, thus generating a set of steep change points. The region generation submodule calls the abrupt change point set, sets the neighborhood radius threshold based on the distribution of measurement point spacing, performs nearest neighbor merging operation on other abrupt change points within the search radius of each abrupt change point, forms a connected region after merging, records the measurement point set of each region, and obtains the abrupt change region set. The feature extraction submodule extracts the minimum and maximum coordinates of the measuring points for each region within the abrupt change region set, calculates the width and length, multiplies the width and length to obtain the region area, calculates the average gradient vector direction angle of all measuring points as the gradient direction angle, and uses the area and direction angle to form the feature value of the abrupt change region.

[0012] Preferably, the direction sensitivity determination module includes: The orientation angle sorting submodule receives gradient orientation angle data, sorts the gradient orientation angles in ascending order of value, takes the value at the middle position of the sorted sequence as the deformation principal orientation angle, and obtains the deformation principal orientation angle value. The sensitivity determination submodule obtains the preset assembly direction vector based on the deformation principal direction angle value, calculates the cosine value of the angle between the deformation principal direction angle value and the assembly direction vector. The cosine value is equal to the dot product of the two vectors divided by the product of their magnitudes. The cosine value is compared with 0.8. If the cosine value is greater than or equal to 0.8, the direction sensitivity type value is assigned as 1; otherwise, it is assigned as 0, thus obtaining the direction sensitivity type value.

[0013] Preferably, the spatial autocorrelation analysis module includes: The global statistics calculation submodule obtains the mean of the grid residuals, sums up the mean of the residuals of all grids to get the sum, divides the sum by the total number of grids to get the global mean, subtracts the global mean from the mean of the residuals of each grid to get the difference, squares the difference, sums it up and divides it by the total number of grids, takes the square root of the difference to get the global standard deviation, and generates the global mean and standard deviation. The Moran index filtering submodule calculates the sum of the products of the differences between the residual mean of each grid and the residual mean of its neighboring grids, based on the global mean, standard deviation, and grid residual mean. This product is then divided by the total number of grids, multiplied by the global variance, and multiplied by the spatial weight coefficient to obtain the Moran index value. This value is compared with a preset threshold value, and grids that exceed the threshold value are selected as clustered grids. Eight-neighbor merging is then performed, and blocks with an area smaller than a preset threshold are removed to obtain clustered grid blocks. The saliency determination submodule calculates the mean residual of each cluster based on the clustered grid blocks and the global mean and standard deviation. It subtracts the mean residual from the global mean to obtain the difference. The difference is divided by the global standard deviation to obtain the multiple. The multiple is compared with a preset high-level threshold. Those greater than the threshold are marked as high-level. The multiple is compared with a preset medium-level threshold. Those greater than the medium-level threshold but not greater than the high-level threshold are marked as medium-level. The rest are marked as low-level. This generates clustered blocks and their saliency levels.

[0014] Preferably, the comprehensive level output module includes: The spatial overlap matching submodule, based on the region area, orientation sensitivity type, clusters and their salience level, performs intersection detection on the coordinate range of the clusters and the coordinate range of the region area, calculates the overlap between the two coordinate intervals, assigns defect block labels to regions with an overlap greater than zero, and generates defect blocks. The grading coefficient calculation submodule calls the defect block and compares the result of the cosine value of the angle between the direction sensitivity type and the assembly direction vector. If the cosine value is less than 0.8, the coefficient is 1.2; otherwise, it is 0.5. Based on the cluster block and its significance level, if the difference between the internal residual mean and the global mean exceeds two standard deviations, the coefficient is 1.5; if it exceeds one standard deviation, the coefficient is 1.0; otherwise, the coefficient is 0.5. The area of ​​the region corresponding to the defect block is multiplied by the two coefficients to obtain the weighted score. The grade determination submodule compares the weighted score with thresholds A and B in that order. If the weighted score is greater than or equal to threshold A, it indicates grade C (unqualified). If the weighted score is less than threshold A but greater than or equal to threshold B, it indicates grade B (requires attention). If the weighted score is less than threshold B, it indicates grade A (qualified), and the flatness detection grade is generated.

[0015] Compared with existing technologies, the advantages of this invention are as follows: By dividing the residual of the measuring point into a 10 mm grid and calculating the gradient vector of eight neighborhoods, the spatial direction characteristics of deformation are extracted as eight-dimensional gradient data. Then, abrupt change points are screened using the gradient magnitude and clustered into continuous regions, allowing for the preservation of geometric boundaries in cases of severe local deformation. Furthermore, the median of the gradient direction angle is extracted and the cosine of the assembly direction is calculated. Directional sensitivity is introduced as a weighting coefficient into the grading determination, giving higher weight to deformations developing along the assembly direction and avoiding the neglect of directional defects that significantly affect sealing. The Moran index is used to analyze the spatial autocorrelation of the mean of the grid residuals, identifying weak deformation clusters such as gradual warping. The significance level is determined by comparing the difference between the internal mean and the global mean with the standard deviation, compensating for the omission of spatially related defects by a single threshold. Finally, the area, directional sensitivity weights, and cluster significance weights are multiplied to obtain the grading value, which is compared with a set threshold to output a C, B, or A grade result. This improves flatness detection from single-value judgment to multi-dimensional weighted grading, enhancing the ability to distinguish complex deformation patterns and quantify risks. Attached Figure Description

[0016] The accompanying drawings, which are provided to further illustrate the invention and constitute a part of this invention, are illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention.

[0017] Figure 1 This is a system block diagram of the present invention. Detailed Implementation

[0018] The following will describe in detail the implementation of the present invention with reference to the accompanying drawings and embodiments, so that the process of how the present invention uses technical means to solve technical problems and achieve technical effects can be fully understood and implemented accordingly.

[0019] In the description of this invention, it should be noted that the directional terms such as "center", "lateral", "longitudinal", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", and "counterclockwise" indicate the orientation and positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. They should not be construed as limiting the specific protection scope of this invention.

[0020] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features. Thus, the use of "first" and "second" to define a feature may explicitly or implicitly include one or more of that feature, and in the description of this invention, "a number" means two or more, unless otherwise explicitly specified.

[0021] In this invention, unless otherwise explicitly specified and limited, the terms "assembly," "connection," and "joining" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can also refer to a mechanical connection; they can refer to a direct connection or a connection through an intermediate medium; or they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0022] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0023] It should also be understood that the terminology used in this specification of embodiments of the invention is for the purpose of describing particular embodiments only and is not intended to limit the embodiments of the invention. As used in this specification of embodiments of the invention and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0024] This embodiment discloses a cylinder head flatness detection device, including a flatness detection system, the detection system comprising: The residual gradient construction module is used to obtain the residual value of the measuring point, divide it into a 10mm×10mm grid and calculate the mean value of the grid residual. For each measuring point, the eight-neighbor residual difference operation is performed to obtain an eight-dimensional vector as the measuring point gradient vector. The mean value of the grid residual is passed to the spatial autocorrelation analysis module and the measuring point gradient vector is passed to the steep change region identification module. The abrupt change region identification module is used to extract the magnitude based on the gradient vector of the measurement point and compare it with the threshold to filter abrupt change points. It uses the nearest neighbor clustering algorithm to cluster abrupt change region, calculates the region area and extracts the gradient direction angle, and passes the region area to the comprehensive level output module and the gradient direction angle to the direction sensitivity determination module. The orientation sensitivity determination module is used to obtain the main deformation direction based on the median of the gradient orientation angle in ascending order, calculate the cosine value of the angle between the orientation and the assembly orientation vector and compare it with 0.8 to generate the orientation sensitivity type, and then pass the orientation sensitivity type to the comprehensive level output module. The spatial autocorrelation analysis module is used to calculate the global mean and standard deviation based on the mean of grid residuals, use Moran's index to obtain statistics to compare critical values ​​and filter clustered grids, merge and remove small blocks, calculate the difference between the internal residual mean and the global mean, compare it with the standard deviation multiple to obtain clustered blocks and their significance levels, and pass the clustered blocks and their significance levels to the comprehensive level output module. The comprehensive grade output module receives the area, orientation sensitivity type, and clusters and their salience levels. It generates defect blocks based on spatial overlap matching. The coefficient is 1.2 or 0.5 based on the orientation sensitivity type, and 1.5, 1.0, or 0.5 based on the clusters and their salience levels. The area is multiplied by the two coefficients to obtain the grade value, which is compared with the thresholds A and B to obtain the flatness detection grade, i.e., grade C is unqualified, grade B requires attention, or grade A is qualified, which is the system output result.

[0025] The mean residual of the grid includes local warping parameters and unit surface undulation; the gradient vector of the measuring point includes microscopic physical potential and microstructural skewness; the area of ​​the region includes physical sweep range and geometric coverage scale; the gradient direction angle includes tilt azimuth parameters and morphological deflection attributes; the direction sensitivity type includes structural interlocking influence category and mechanical docking obstacle qualitative; and the clusters and their significance levels include contiguous mutation groups and local deterioration severity assessment conclusions.

[0026] in,

[0027] Local warpage parameter: Characterizes the curvature change in a certain area, usually obtained by laser measurement or contact measurement, and measures the degree of warpage on the cylinder head surface.

[0028] Unit surface roughness: Measures the change in height of a surface within a unit area, usually detected using surface roughness instruments or optical methods.

[0029] Microscopic physical high and low potentials: describe the changes in height at the physical level on the surface of a microstructure, usually involving the application of microscopic techniques.

[0030] Minor structural skewness: Measures the degree of minute deviation from the surface of a structure, which can be evaluated using precision measuring equipment.

[0031] Physical impact range: refers to the physical spatial range affected by a specific defect, which is usually obtained through data analysis and model derivation.

[0032] Geometric coverage scale: describes the extent to which a defect extends within geometric space, typically involving how surface area or volume is calculated.

[0033] Tilt orientation parameter: refers to the measurement of the tilt angle of the surface, which is usually obtained through an angle measuring instrument.

[0034] Morphological deflection attribute: Describes the deformation and deflection characteristics of the object's surface morphology, derived by comparing an ideal model with actual measurement data.

[0035] Structural interlocking effect category: This category includes interlocking effects that may occur when structural connections are made. Potential effects are caused by material properties and design flaws.

[0036] Qualitative assessment of mechanical docking obstacles: Assessing the obstacles encountered during mechanical installation or docking and their impact.

[0037] Clustered mutation groups: describe a group of adjacent defect points that often affect the overall performance of the structure together.

[0038] Localized deterioration severity assessment conclusion: The judgment on the performance deterioration of a local area is derived by comprehensively considering multiple detection indicators.

[0039] For example, the residual gradient building module includes: The numerical acquisition submodule obtains the residual values ​​of the measuring points. First, 100 measuring points are evenly distributed on the plane to be inspected on the cylinder head. The coordinates of the measuring points are pre-calibrated by a coordinate measuring machine; for example, measuring points are numbered 1 to 100, and their plane coordinates (X...)... i , Y i The unit is millimeters, for example, X1=0, Y1=0, X2=10, Y2=0, up to X 10 =90、Y 10 =0, second row Y 11 =10, and so on to form a 10×10 grid. At each measuring point, a laser displacement sensor is used to collect the actual height H. 实测,i Simultaneously, the theoretical height H of this point is extracted from the cylinder head CAD model. 理论,i Subtracting the two yields the residual value R. i = H 实测,i - H 理论,i For example, if the measured height of measuring point 1 is 10.02 mm and the theoretical height is 10.00 mm, then R1 = 0.02 mm; if the measured height of measuring point 2 is 9.98 mm and the theoretical height is 10.00 mm, then R2 = -0.02 mm. The residual values ​​for all measuring points are recorded in Table 1. When extracting the residual value from the measurement data for each measuring point, R is read from the sensor output file. i And verify its sign and range, for example, R i The range is from -0.05 mm to 0.05 mm; values ​​exceeding this range are marked as abnormal and discarded. When associating the residual values ​​with the plane coordinates of the measuring points, a data structure such as an array [X] is established. i , Y i , R i For example, measurement point 1 corresponds to [0, 0, 0.02], and measurement point 2 corresponds to [10, 0, -0.02]. This ensures that each coordinate uniquely corresponds to a residual value, ultimately resulting in a dataset of measurement point residual values ​​containing 100 sets of coordinate-residual mapping relationships. Table 1 lists some of the measurement point residual value data. As shown in Table 1, a positive residual value indicates that the measured height is higher than the theoretical height, and a negative value indicates that it is lower than the theoretical height. Subsequent modules will perform grid division and gradient calculation based on these associated data.

[0040] Table 1. Residual values ​​of some measuring points , The block averaging submodule, based on the residual values ​​of the measuring points, divides the measuring point plane into square grids with a side length of 10 mm. First, it takes the minimum x-coordinate value of 0 mm and the maximum x-coordinate value of 90 mm for all measuring points, and the minimum y-coordinate value of 0 mm and the maximum y-coordinate value of 90 mm. Starting from the origin, it generates grid lines with a step size of 10 mm, where the horizontal lines are... The vertical line is ,form There are 100 grids in total, and each grid is numbered as shown in grid (1,1). The grid (1,2) corresponds to Similarly, the residual values ​​of all measuring points within each grid are summed and divided by the number of measuring points in that grid. For example, grid (1,1) contains measuring points 1, 2, 11, and 12, a total of 4 measuring points, with residual values ​​of 0.02, 0.01, -0.01, and 0.03 mm respectively. The summation yields... If the measurement point is 4 millimeters, then the mean grid residual is _____. For edge grids, the number of measurement points may be insufficient. For example, if grid (10,10) only contains measurement point 100, the mean value is equal to the residual value of that point, which is 0.02 mm. Iterate through all 100 grids and record the mean residual value for each grid. ,in , The formula used in calculating the mean residual of this grid is: ,in This represents the number of measurement points within the grid. These represent the residual values ​​at the measuring points. Taking grid (2,3) as an example, the measured residuals are -0.02, 0.04, 0.01, and -0.03 mm, respectively. Summing the results gives 0.00 millimeters. Millimeters. Global mean and standard deviation In the statistics, we assume there are 100 items. The total is 1.25 mm, then Millimeters; each grid and Assuming the sum of squared differences is 0.035 square millimeters, then Millimeters. The advantage of this grid residual mean formula is that it eliminates single-point noise through grid averaging, making local region characteristics more stable and providing a basis for subsequent spatial autocorrelation analysis. Table 2 shows some grid residual mean data, and provides the test results of the embodiment. This value is directly used for Moran's index calculation.

[0041] Table 2. Mean Residual Data for Partial Grid Grids , The vector construction submodule, for each measurement point, calls the measurement point residual value and extracts the measurement point residual values ​​at eight positions: up, down, left, right, and four diagonal directions. Subtracting the measurement point residual value from each neighboring residual value yields eight difference values. These eight differences are arranged into a vector in clockwise order starting from due north, resulting in the measurement point gradient vector. First, taking measurement point 1 (coordinates 0,0) as an example, its residual... Millimeters, the measuring point is (0,10) in the due north direction, which is also the measuring point 11, residual millimeters, difference Millimeters, the measuring point is (10,0) in the due east direction, which is measuring point 2, residual millimeters, difference The measurement point is located in millimeters. The south direction is at point (0, -10), but this coordinate system does not exist outside the plane, so it is marked as missing. The west direction is also missing. The northeast direction has a measurement point (10, 10), which is measurement point 12. The residual... millimeters, difference The measurement point (10, -10) in the southeast direction does not exist, nor does it exist in the southwest or northwest directions. For boundary measurement points, missing differences in the eight neighborhoods are filled with 0. Therefore, the eight neighborhood difference sequence of measurement point 1 is: North 0.01, Northeast -0.01, East -0.04, Southeast 0, South 0, Southwest 0, West 0, Northwest 0. Arranging these in clockwise order yields the gradient vector. The unit is millimeters. For an internal measuring point such as measuring point 22 (coordinates 10, 10), all eight of its neighborhoods exist. For example, the residual of the north measuring point (10, 20) is 0.02, the residual of the east measuring point (20, 10) is 0.01, the residual of the south measuring point (10, 0) is -0.01, the residual of the west measuring point (0, 10) is 0.03, the residual of the northeast measuring point (20, 20) is 0.04, the residual of the southeast measuring point (20, 0) is -0.02, the residual of the southwest measuring point (0, 0) is 0.02, and the residual of the northwest measuring point (0, 20) is 0.01. The residual of this measuring point itself is 0.01 millimeters. Therefore, the complete vector is obtained by calculating the eight differences. The formula for calculating each difference is as follows: ,in This indicates one of eight directions. Taking measuring point 22 as an example... , The advantage of this formula lies in capturing the anisotropy of local deformation through eight-directional difference, providing directional information for subsequent identification of abrupt change points.

[0042] For example, the abrupt change region identification module includes: The steep change point filtering submodule calculates the Euclidean magnitude of each gradient vector based on the measurement point's gradient vector. It then compares each magnitude with a set magnitude threshold. Measurement points with magnitudes exceeding the threshold are marked as steep changes, generating a set of steep change points. First, the magnitude of the gradient vector at each measurement point is calculated. For example, the vector of measurement point 1 The sum of their squares is , Module length Millimeters, measuring point 22 vector The sum of squares is , Module length Millimeters. The modulus threshold is set based on the statistical distribution: calculate the average modulus of all 100 measuring points. millimeters, standard deviation millimeters, take Millimeters are used as a threshold. For example, the modulus of measuring point 50. For measurements greater than 0.075 mm, measurement point 50 is marked as abruptly changing. The advantage of this modulus formula is that it quantifies the severity of local fluctuations using the Euclidean norm and uses a threshold of 2 standard deviations to screen out significant outliers. The final abruptly changing point set contains 15 measurement points and is output to the region generation submodule.

[0043] The region generation submodule calls the abrupt change point set, sets a neighborhood radius threshold based on the distribution of measurement point spacing, and performs a nearest neighbor merging operation on other abrupt change points within the search radius of each abrupt change point. After merging, a connected region is formed, and the measurement point sets of each region are recorded to obtain the abrupt change region set. Initially, the measurement point spacing is 10 mm (grid step size), and the neighborhood radius threshold is set to 15 mm, meaning it covers at least one adjacent measurement point. A point is randomly selected from the abrupt change point set, such as measurement point 50, with coordinates (50, 50). Other abrupt change points within a 15 mm search radius are also considered, for example, measurement point 48 (40, 50) is 10 mm away, measurement point 52 (60, 50) is 10 mm away, and measurement point 49 (50, 40) is 10 mm away. All meet the condition, so these points are merged into the same region, and the process continues with a new... The merged points are searched repeatedly until no new points are added. For example, region A is formed, containing measurement points 48, 49, 50, and 52. At the same time, measurement point 68 (70, 80) is near measurement point 67 (70, 70) at a distance of 10 mm, forming region B, which contains measurement points 67 and 68. The remaining isolated abrupt change points, such as measurement point 73 (90, 90), have no other abrupt change points nearby and form a separate region C. Finally, three abrupt change regions are obtained. The coordinate set of measurement points in each region is recorded and output to the feature extraction submodule.

[0044] The feature extraction submodule extracts the minimum and maximum coordinates of the measuring points for each region within the abrupt change region set, calculates the width and length, multiplies the width and length to obtain the region area, and calculates the mean gradient vector direction angle of all measuring points as the gradient direction angle. The area and direction angle together form the feature value of the abrupt change region. For example, region A contains measuring points 48(40,50), 49(50,40), 50(50,50), and 52(60,50), with a minimum x-coordinate of 40 mm, a maximum x-coordinate of 60 mm, a width of 20 mm, a minimum y-coordinate of 40 mm, a maximum y-coordinate of 50 mm, a length of 10 mm, and an area of... Square millimeters. Calculate the gradient vector direction angle at each measuring point, simplifying the process by taking the direction corresponding to the largest component of the vector as the principal direction. For example, for measuring point 50, vector... The maximum component is 0.03, corresponding to the due east direction (the third position), and the angle is defined as... Then the direction angle is The maximum component of the measuring point 48 corresponds to the angle in the true north direction (first position). The maximum component at measuring point 49 corresponds to true north. The maximum component at measuring point 52 corresponds to due east. mean gradient direction angle of region A Similarly, region B has an area of ​​150 square millimeters and an orientation angle of... Region C has an area of ​​100 square millimeters and an orientation angle of... These feature values ​​are passed to the orientation sensitivity determination module and the comprehensive level output module.

[0045] For example, the orientation sensitivity determination module includes: The orientation angle sorting submodule receives gradient orientation angle data, sorts the gradient orientation angles in ascending order of value, and takes the value at the middle position of the sorted sequence as the principal deformation orientation angle. For example, the gradient orientation angles of the three regions are... After sorting, it is The median is (If there are an odd number of values, take the middle value), then the principal deformation direction angle If the number of regions is even, the average of the two middle numbers is taken and output to the sensitivity determination submodule.

[0046] The sensitivity determination submodule, based on the deformation principal direction angle value, obtains a preset assembly direction vector, calculates the cosine of the angle between the deformation principal direction angle value and the assembly direction vector, and the cosine value is equal to the dot product of the two vectors divided by the product of their magnitudes. The cosine value is compared with 0.8; if the cosine value is greater than or equal to 0.8, the direction sensitivity type value is assigned 1; otherwise, it is assigned 0. Firstly, the assembly direction vector is determined by the design drawings; for example, the direction of the bolt hole axis during cylinder head installation is... Corresponding unit vector principal direction angle of deformation Corresponding unit vector According to the formula dot product The modulus length is 1, therefore Since the value is less than 0.8, the direction sensitivity type value is assigned to 0, indicating that the deformation direction is inconsistent with the assembly direction, affecting the structural engagement. If the main deformation direction is... ,but The dot product is 0.9848, and the cosine value of 0.9848 is greater than 0.8, so the type value is assigned to 1. The advantage of this formula is that it quantitatively evaluates the degree to which the deformation direction hinders assembly through the cosine value; a threshold of 0.8 corresponds to an angle of approximately 36.87°. A cosine value of 0.7071 indicates an angle of 45°, with a direction sensitivity type of 0, leading to a subsequent coefficient of 1.2, which exacerbates the defect level.

[0047] For example, the spatial autocorrelation analysis module includes: The global statistics calculation submodule obtains the mean of the grid residuals, sums all the mean residuals of the grids to get the sum, divides the sum by the total number of grids to get the global mean, subtracts the global mean from the mean of the mean residuals of each grid to get the difference, squares the difference, sums them, divides them by the total number of grids, and takes the square root of the difference to get the global standard deviation. Initially, it is assumed that the sum of the mean residuals of all 100 grids is 1.25 mm, and the total number of grids... According to the formula Calculate the global mean Millimeters. Calculate the difference between each grid cell. For example, the interpolation of grid (1,1) Grid (1,2) difference The sum of the squares is 0.035 square millimeters. According to the formula... Calculated Millimeters. This result reflects the degree of dispersion of the overall fluctuations. Millimeters indicate a slightly higher overall value. Millimeters are used as a benchmark for subsequent significance factor calculations.

[0048] The Moran's index filtering submodule, based on the global mean, standard deviation, and grid residual mean, calculates the sum of the products of the differences between the residual mean of each grid and the mean of the residual of its neighboring grids. This sum is then divided by the total number of grids, multiplied by the global variance, and multiplied by the spatial weighting coefficient to obtain the Moran's index value. This value is compared with a preset threshold, and grids exceeding the threshold are selected as clustered grids. Eight-neighbor merging is then performed, and blocks with areas smaller than a preset threshold are removed, resulting in clustered grid blocks. The Moran's index formula is as follows: ,in , This is the spatial adjacency weight (1 for eight-neighborhood adjacency, 0 otherwise). Square millimeters. Assuming a grid of (5,5) The value is in millimeters, and the mean residuals of its eight neighborhoods are 0.030, 0.032, 0.028, 0.040, 0.036, 0.034, 0.038, and 0.042, respectively. Calculate its local molecules: The sum of the products of the product and the differences of each neighborhood is assumed to be 0.045. The numerator sum is calculated similarly for all grids. =3.5 (illustrative value), denominator remains unchanged. Weighted sum For 100 grids, the estimate is 540. (The indicated value is too large; the actual value should be smaller.) To satisfy the calculation example, reasonable assumptions are made: the actual calculated value is... The value is greater than the critical value of 0.2, so the grid (5,5) is marked as a clustered grid. After sequential filtering, the clustered grids are merged into eight-neighborhoods to form clustered blocks. Blocks with an area less than 500 square millimeters are removed, and finally the clustered blocks and their coordinate ranges are obtained. The advantage of this formula is that it can quantitatively identify the spatial clustering of residuals; non-randomly distributed areas are the true contiguous anomalies.

[0049] The significance determination submodule, based on clustered grid blocks and the global mean and standard deviation, calculates the mean residual within each clustered block, subtracts it from the global mean to obtain the difference, and divides the difference by the global standard deviation to obtain the multiplier. The multiplier is compared with the preset high-level threshold of 2.0 and the medium-level threshold of 1.0. For example, if cluster D contains grids (5,5), (5,6), and (6,5), and their internal residual mean values ​​are 0.035, 0.042, and 0.038 mm, respectively, then... millimeters The difference is 0.02583 mm. millimeters .because This is considered to be of medium significance. That would be a higher level. This is a low-level formula. Its advantage lies in eliminating the effects of global fluctuations through standardization factors, making data from different batches comparable. This indicates that the cluster deviates from the global mean by more than one standard deviation but less than two, which is considered a moderate deterioration.

[0050] For example, the comprehensive rating output module includes: The spatial overlap matching submodule, based on region area, orientation sensitivity type, and clusters and their salience levels, performs intersection detection on the coordinate ranges of clusters and region areas, calculates the overlap between their coordinate intervals, and assigns defect block labels to regions with an overlap greater than zero. For example, the coordinate range of abruptly changing region A is 40-60 mm x-axis and 40-50 mm y-axis, and the coordinate range of cluster D is also 40-60 mm x-axis and 40-60 mm y-axis. The overlap area... Since the square millimeter value is greater than zero, region A overlaps with cluster D and is therefore identified as a defective block. Similarly, regions B and C, which do not overlap with clusters, are not identified. Finally, a list of defective blocks is generated and output to the grading coefficient calculation submodule.

[0051] The submodule for calculating the grading coefficient calls the defect block. Based on the comparison result between the direction sensitivity type and the cosine value of the angle between the assembly direction vector and the defect block, a coefficient of 1.2 is taken if the cosine value is less than 0.8, otherwise 0.5 is taken. Based on the clustered blocks and their significance levels, a coefficient of 1.5 is taken if the difference between the internal residual mean and the global mean exceeds two standard deviations, a coefficient of 1.0 is taken if the difference exceeds one standard deviation, and a coefficient of 0.5 is taken otherwise. The area of ​​the region corresponding to the defect block is multiplied by the two coefficients to obtain the weighted score. For example, a defective block consists of region A and cluster D, with a direction sensitivity type value of 0 (cosine 0.7071 is less than 0.8). The significance level of the cluster is medium (multiple 1.381). area square millimeters, then If the orientation sensitivity type value is 1, then If the significance level is high, then lower level The advantage of this formula is that it integrates three independent dimensions—area, orientation, and salience—and uses weighted coefficients to differentiate scores for defects of varying severity.

[0052] The rating determination submodule will assign weighted scores. The decision rule is based on the order of execution of thresholds A and B: if... Then it is grade C and unqualified. Then it is classified as Grade B and needs attention. This is considered a Grade A pass. Thresholds A and B are set based on historical data statistics: 50 qualified cylinder heads are tested. All are below 100, so set ; 10 defective products were inspected. All are above 300, so set This example ,satisfy It is classified as Grade B and requires attention. The advantage of this judgment logic is that the grading threshold is set based on actual data, making the assessment results objective and repeatable. This indicates that the area needs further observation or minor processing, and the output is directly used as the system flatness test level.

[0053] The above description only illustrates the preferred embodiments of the present invention and should not be construed as limiting the scope of the claims. The present invention is not limited to the above embodiments, and variations in its specific structure are permitted. All modifications made within the scope of the independent claims of this invention are also within the scope of protection of this invention.

Claims

1. A cylinder head flatness testing device, characterized in that, Includes a flatness detection system, the detection system comprising: The residual gradient construction module is used to obtain the residual value of the measuring point, divide it into a 10mm×10mm grid and calculate the mean value of the grid residual. For each measuring point, the eight-neighbor residual difference operation is performed to obtain an eight-dimensional vector as the measuring point gradient vector. The mean value of the grid residual is passed to the spatial autocorrelation analysis module and the measuring point gradient vector is passed to the steep change region identification module. The abrupt change region identification module is used to extract the magnitude based on the gradient vector of the measurement point and compare it with the threshold to filter abrupt change points. It uses the nearest neighbor clustering algorithm to cluster abrupt change region, calculates the region area and extracts the gradient direction angle, and passes the region area to the comprehensive level output module and the gradient direction angle to the direction sensitivity determination module. The orientation sensitivity determination module is used to obtain the main deformation direction based on the median of the gradient orientation angle in ascending order, calculate the cosine value of the angle between the orientation and the assembly orientation vector and compare it with 0.8 to generate the orientation sensitivity type, and then pass the orientation sensitivity type to the comprehensive level output module. The spatial autocorrelation analysis module is used to calculate the global mean and standard deviation based on the mean of grid residuals, use Moran's index to obtain statistics to compare critical values ​​and filter clustered grids, merge and remove small blocks, calculate the difference between the internal residual mean and the global mean, compare it with the standard deviation multiple to obtain clustered blocks and their significance levels, and pass the clustered blocks and their significance levels to the comprehensive level output module.

2. The cylinder head flatness detection device according to claim 1, characterized in that, It also includes a comprehensive grade output module, which receives the area, orientation sensitivity type, and clusters and their salience levels. It generates defect blocks by matching spatial overlap. The coefficient is 1.2 or 0.5 based on the orientation sensitivity type, and 1.5, 1.0, or 0.5 based on the clusters and their salience levels. The area is multiplied by the two coefficients to obtain the grade value, which is compared with the thresholds A and B to obtain the flatness detection grade, i.e., grade C is unqualified, grade B requires attention, or grade A is qualified, which is the system output result.

3. The cylinder head flatness detection device according to claim 1, characterized in that, The mean residual of the grid includes local warping parameters and unit surface undulation, and the gradient vector of the measurement point includes microscopic physical potential and microstructural skewness.

4. The cylinder head flatness detection device according to claim 1, characterized in that, The area of ​​the region includes the physical range of impact and the geometric coverage, and the gradient direction angle includes the tilt azimuth parameter and the shape deflection attribute.

5. The cylinder head flatness testing device according to claim 1, characterized in that, The directional sensitivity types include structural interlocking influence categories and mechanical docking obstacle qualitative analysis. The aggregates and their significance levels include contiguous mutation groups and local deterioration severity assessment conclusions.

6. The cylinder head flatness testing device according to claim 1, characterized in that, The residual gradient construction module includes: The numerical acquisition submodule obtains the residual values ​​of the measuring points, extracts the residual values ​​from the measurement data for each measuring point, and associates these values ​​with the plane coordinates of the measuring point to obtain the residual values ​​of the measuring points. The block mean submodule divides the measuring point plane into square grids with a side length of 10 mm based on the residual values ​​of the measuring points. It sums up the residual values ​​of all measuring points in each grid and divides them by the number of measuring points in that grid to obtain the grid mean residual value. The vector construction submodule, for each measurement point, calls the measurement point residual value and extracts the measurement point residual values ​​at eight positions: up, down, left, right, and four diagonal directions. Subtracting the measurement point residual value from each neighborhood residual value yields eight difference values. Arrange the eight difference values ​​into a vector in clockwise order starting from due north to obtain the measurement point gradient vector.

7. The cylinder head flatness testing device according to claim 1, characterized in that, The abrupt change region identification module includes: The steep change point filtering submodule calculates the Euclidean magnitude of each vector based on the gradient vector of the measurement point, compares each magnitude with a set magnitude threshold, and assigns steep change markers to measurement points whose magnitudes exceed the threshold, thus generating a set of steep change points. The region generation submodule calls the abrupt change point set, sets the neighborhood radius threshold based on the distribution of measurement point spacing, performs nearest neighbor merging operation on other abrupt change points within the search radius of each abrupt change point, forms a connected region after merging, records the measurement point set of each region, and obtains the abrupt change region set. The feature extraction submodule extracts the minimum and maximum coordinates of the measuring points for each region within the abrupt change region set, calculates the width and length, multiplies the width and length to obtain the region area, calculates the average gradient vector direction angle of all measuring points as the gradient direction angle, and uses the area and direction angle to form the feature value of the abrupt change region.

8. The cylinder head flatness testing device according to claim 1, characterized in that, The orientation sensitivity determination module includes: The orientation angle sorting submodule receives gradient orientation angle data, sorts the gradient orientation angles in ascending order of value, takes the value at the middle position of the sorted sequence as the deformation principal orientation angle, and obtains the deformation principal orientation angle value. The sensitivity determination submodule obtains the preset assembly direction vector based on the deformation principal direction angle value, calculates the cosine value of the angle between the deformation principal direction angle value and the assembly direction vector. The cosine value is equal to the dot product of the two vectors divided by the product of their magnitudes. The cosine value is compared with 0.

8. If the cosine value is greater than or equal to 0.8, the direction sensitivity type value is assigned as 1; otherwise, it is assigned as 0, thus obtaining the direction sensitivity type value.

9. A cylinder head flatness testing device according to claim 1, characterized in that, The spatial autocorrelation analysis module includes: The global statistics calculation submodule obtains the mean of the grid residuals, sums up the mean of the residuals of all grids to get the sum, divides the sum by the total number of grids to get the global mean, subtracts the global mean from the mean of the residuals of each grid to get the difference, squares the difference, sums it up and divides it by the total number of grids, takes the square root of the difference to get the global standard deviation, and generates the global mean and standard deviation. The Moran index filtering submodule calculates the sum of the products of the differences between the residual mean of each grid and the residual mean of its neighboring grids, based on the global mean, standard deviation, and grid residual mean. This product is then divided by the total number of grids, multiplied by the global variance, and multiplied by the spatial weight coefficient to obtain the Moran index value. This value is compared with a preset threshold value, and grids that exceed the threshold value are selected as clustered grids. Eight-neighbor merging is then performed, and blocks with an area smaller than a preset threshold are removed to obtain clustered grid blocks. The saliency determination submodule calculates the mean residual of each cluster based on the clustered grid blocks and the global mean and standard deviation. It subtracts the mean residual from the global mean to obtain the difference. The difference is divided by the global standard deviation to obtain the multiple. The multiple is compared with a preset high-level threshold. Those greater than the threshold are marked as high-level. The multiple is compared with a preset medium-level threshold. Those greater than the medium-level threshold but not greater than the high-level threshold are marked as medium-level. The rest are marked as low-level. This generates clustered blocks and their saliency levels.

10. A cylinder head flatness testing device according to claim 2, characterized in that, The comprehensive rating output module includes: The spatial overlap matching submodule, based on the region area, orientation sensitivity type, clusters and their salience level, performs intersection detection on the coordinate range of the clusters and the coordinate range of the region area, calculates the overlap between the two coordinate intervals, assigns defect block labels to regions with an overlap greater than zero, and generates defect blocks. The grading coefficient calculation submodule calls the defect block and compares the result of the cosine value of the angle between the direction sensitivity type and the assembly direction vector. If the cosine value is less than 0.8, the coefficient is 1.2; otherwise, it is 0.

5. Based on the cluster block and its significance level, if the difference between the internal residual mean and the global mean exceeds two standard deviations, the coefficient is 1.5; if it exceeds one standard deviation, the coefficient is 1.0; otherwise, the coefficient is 0.

5. The area of ​​the region corresponding to the defect block is multiplied by the two coefficients to obtain the weighted score. The grade determination submodule compares the weighted score with thresholds A and B in that order. If the weighted score is greater than or equal to threshold A, it indicates grade C (unqualified). If the weighted score is less than threshold A but greater than or equal to threshold B, it indicates grade B (requires attention). If the weighted score is less than threshold B, it indicates grade A (qualified), and the flatness detection grade is generated.

Images