A non-destructive testing method, testing device and system for surface hardness of a steel ball

Abstract

The present application relates to the technical field of nondestructive testing of steel balls, and particularly relates to a nondestructive testing method, a testing device and a system for surface hardness of a steel ball. The method comprises: obtaining initial hardness testing data and spatial position information of a plurality of testing points; dividing the testing points into corresponding spatial annular zones according to the spatial position information; dividing the testing points into a standard testing group and an abnormal testing point set based on the initial hardness testing data in each spatial annular zone; calculating a radius modulus difference degree between the standard testing group and the abnormal testing point set, and analyzing the radius modulus difference degree to obtain abnormal testing point heterogeneity; determining a weight coefficient of the abnormal testing point set according to the abnormal testing point heterogeneity; correcting the initial hardness testing data of the abnormal testing point set based on the weight coefficient to obtain corrected hardness testing data, and obtaining a target hardness testing result based on the corrected hardness testing data. The present application can improve the nondestructive testing effect of the surface hardness of the steel ball.

Description

Technical Field

[0001] This invention relates to the field of nondestructive testing technology for steel balls, specifically to a nondestructive testing method, testing equipment, and system for the surface hardness of steel balls. Background Technology

[0002] In automated production lines for continuous heat treatment of wear-resistant steel balls, surface hardness uniformity is a key quality control indicator for ensuring the impact resistance, fracture resistance, and wear resistance performance of the steel balls. Traditional hardness testing methods mainly rely on offline indentation sampling in the laboratory and cross-sectional inspection of the steel balls. Both of these methods cause irreversible damage to the steel balls and can only select local points for testing, failing to completely cover the entire spherical surface of the steel ball. This easily creates blind spots in the testing, leading to the omission of workpieces with local hardness defects.

[0003] Currently, electromagnetic eddy current nondestructive testing (EDT) technology has been gradually applied to online production line inspection of steel balls. This technology relies on the principle of electromagnetic induction to complete the inspection without damaging the steel ball matrix, eliminating the workpiece loss and production cost losses caused by traditional indentation and sectioning inspections. However, existing eddy current-supported hardness data evaluation algorithms cannot effectively decouple qualified and abnormal test data: qualified data corresponds to electromagnetic response signals output from measurement points with intact metallographic structure and uniform hardness across the entire surface of the steel ball, while abnormal data originates from test signals generated in areas of localized metallographic deterioration and hardness reduction caused by fluctuations in heat treatment processes and surface decarburization. Abnormal data directly participates in the overall hardness calculation, ultimately resulting in distorted hardness evaluation results and low reliability, which cannot meet the requirements of modern steel ball production lines for high-precision and high-reliability online hardness testing of the entire surface.

[0004] In other words, the existing technology for non-destructive testing of the surface hardness of steel balls is not very effective. Summary of the Invention

[0005] The purpose of this invention is to provide a non-destructive testing method, testing equipment, and system for the surface hardness of steel balls, in order to solve the technical problem that the non-destructive testing effect of steel ball surface hardness is poor in the prior art.

[0006] In a first aspect, one embodiment of the present invention provides a non-destructive testing method for the surface hardness of a steel ball, the method comprising: The initial hardness detection data and corresponding spatial location information of multiple detection points on the surface of the steel ball to be tested are obtained; the initial hardness detection data includes complex impedance eigenvectors; the complex impedance eigenvectors are used to characterize the electromagnetic response characteristics of the steel ball at the detection points; Based on spatial location information, multiple detection points are divided into corresponding spatial rings. Within each spatial ring, based on the distribution characteristics of the initial hardness test data, multiple detection points are divided into standard test groups and abnormal test point sets. The spatial rings are used to divide the surface of the steel ball to be tested into multiple analysis regions radially. The radius modulus difference between the standard detection group and the abnormal detection point set is calculated, and the radius modulus difference is analyzed to obtain the heterogeneity of the abnormal detection points. The radius modulus difference is used to characterize the degree of deviation between the electromagnetic response characteristics of the abnormal detection point set and the standard detection group as a function of radius. The heterogeneity of the abnormal detection points is used to characterize the deviation stability of the electromagnetic response behavior of the abnormal detection point set relative to the standard detection group during the detection process. Based on the heterogeneity of anomaly detection points, determine the weight coefficients corresponding to the anomaly detection point set; The initial hardness test data of the abnormal detection point set is corrected based on the weighting coefficient to obtain the corrected hardness test data. The corrected hardness test data is then weighted and fused with the initial hardness test data of the standard test group to obtain the target hardness test result.

[0007] In one embodiment, the step of dividing multiple test points into a standard test group and an abnormal test point set based on the distribution characteristics of the initial hardness test data within each spatial annulus includes: For each spatial ring, calculate the mean and standard deviation of the initial hardness test data for all test points within the spatial ring; Based on the mean and standard deviation, determine the range of detection data deviation corresponding to the spatial ring zone; Test points whose initial hardness test data are within the test data deviation range are classified into standard test groups, and test points whose initial hardness test data are not within the test data deviation range are classified into abnormal test point sets.

[0008] In one embodiment, the step of calculating the radius modulus difference between the standard detection group and the anomaly detection point set includes: Calculate the asynchronous degree of the first radius modulus corresponding to the standard detection group and the asynchronous degree of the second radius modulus corresponding to the abnormal detection point set, respectively; The radius modulus difference is obtained based on the absolute value of the difference between the first radius modulus asynchrony degree and the second radius modulus asynchrony degree.

[0009] In one embodiment, the step of analyzing the radius modulus difference to obtain the heterogeneity of abnormal detection points includes: Acquire time-series detection data on the variation of radius modulus difference over time; Based on the detection time series data, the fluctuation residual corresponding to the radius modulus difference is calculated; the fluctuation residual is used to characterize the instability of the radius modulus difference. The uniformity of anomalous data in the set of anomaly detection points is calculated relative to the baseline anomaly detection points. The baseline anomaly detection points are selected from the set of anomaly detection points during the initial detection phase to provide data reference. The uniformity of anomalous data is used to characterize the consistency of electromagnetic response characteristics within the set of anomaly detection points. Based on the fluctuation residuals and the uniformity of outlier data, the heterogeneity of outlier detection points is calculated.

[0010] In one embodiment, the step of determining the weight coefficients corresponding to the set of anomaly detection points based on the heterogeneity of the anomaly detection points includes: The weight adjustment coefficient is calculated based on the heterogeneity of anomaly detection points and the preset threshold. Multiply the initial evaluation weights corresponding to the anomaly detection point set by the weight adjustment coefficients to obtain the weight coefficients corresponding to the anomaly detection point set.

[0011] In one embodiment, the step of weightedly fusing the corrected hardness test data with the initial hardness test data of the standard test group to obtain the target hardness test result includes: The initial hardness test data and the corrected hardness test data of the standard test group are input into the preset hardness inversion model to obtain the corresponding hardness inversion values. The target hardness test result is obtained by weighted averaging of all hardness inversion values.

[0012] In one embodiment, the calculation steps for the asynchronous degree of the first radius modulus corresponding to the standard detection group include: Obtain the radius of the first cross-section circle and the magnitude of the first vector corresponding to each detection point in the standard detection group; The first cross-sectional circle radius and the first vector magnitude are serialized to obtain the scan radius sequence and the vector magnitude sequence, respectively. Calculate the first morphological correlation coefficient between the scan radius sequence and the vector magnitude sequence; Based on the correlation coefficient of the first morphology, the asynchronous degree of the first radius modulus is calculated.

[0013] In one embodiment, the calculation steps for the asynchronous degree of the second radius modulus corresponding to the anomaly detection point set include: Obtain the radius of the second cross-section circle and the magnitude of the second vector corresponding to each detection point in the anomaly detection point set; The radius of the second cross-section circle and the magnitude of the second vector are serialized to obtain the abnormal radius sequence and the abnormal magnitude sequence, respectively. Calculate the second morphological correlation coefficient between the abnormal radius sequence and the abnormal modulus sequence; Based on the correlation coefficient of the second morphology, the asynchronous degree of the second radius modulus is calculated.

[0014] Secondly, another embodiment of the present invention provides a non-destructive testing system for the surface hardness of a steel ball, the system comprising: The acquisition module is used to acquire the initial hardness detection data and corresponding spatial location information of multiple detection points on the surface of the steel ball to be tested; the initial hardness detection data includes a complex impedance feature vector; the complex impedance feature vector is used to characterize the electromagnetic response characteristics of the steel ball to be tested at the detection point. The partitioning module is used to divide multiple detection points into corresponding spatial rings based on spatial location information, and within each spatial ring, based on the distribution characteristics of the initial hardness test data, divide the multiple detection points into standard detection groups and abnormal detection point sets; the spatial rings are used to divide the surface of the steel ball to be tested into multiple analysis regions radially; The analysis module is used to calculate the radius modulus difference between the standard detection group and the abnormal detection point set, and to analyze the radius modulus difference to obtain the heterogeneity of the abnormal detection points. The radius modulus difference is used to characterize the degree of deviation between the electromagnetic response characteristics of the abnormal detection point set and the standard detection group as a function of radius. The heterogeneity of the abnormal detection points is used to characterize the deviation stability of the electromagnetic response behavior of the abnormal detection point set relative to the standard detection group during the detection process. The determination module is used to determine the weight coefficients corresponding to the set of anomaly detection points based on the heterogeneity of the anomaly detection points. The correction module is used to correct the initial hardness test data of the abnormal detection point set based on the weight coefficient, so as to obtain the corrected hardness test data. The fusion module is used to weight and fuse the corrected hardness test data with the initial hardness test data of the standard test group to obtain the target hardness test result.

[0015] Thirdly, in another embodiment of the present invention, an electronic device is provided, including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method described in the first aspect.

[0016] Fourthly, in another embodiment of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0017] The present invention has the following beneficial effects: This invention utilizes a radial spatial annular zone division strategy on the spherical surface of a steel ball, combined with electromagnetic response data characterized by complex impedance eigenvectors, to distinguish between standard detection groups and abnormal detection point sets. This separates defective test points caused by heat treatment decarburization and low hardness from metallographically acceptable test points, addressing the limitation of conventional algorithms in decoupling normal and abnormal detection data. The deviation of the electromagnetic response from the spherical radius variation is quantified by calculating the difference in radius modulus between the standard and abnormal detection point sets. Then, the heterogeneity of abnormal detection points is solved based on time-series fluctuation characteristics, reflecting the degree of defect signal concealment. Adaptive weighting coefficients are dynamically generated based on the heterogeneity of abnormal detection points, reducing the weighting correction of the original electromagnetic data of abnormal detection points and weakening the interference of distorted data from deteriorated structures on the overall evaluation. Finally, the original data from the standard group and the corrected abnormal data are weighted and used to calculate the target hardness detection result. This not only comprehensively covers the detection information of the entire steel ball surface, avoiding missed detection of local defects, but also eliminates system signal interference caused by the nonlinear attenuation of the spherical scanning linear velocity, thereby improving the non-destructive testing effect of steel ball surface hardness. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a non-destructive testing method for the surface hardness of a steel ball provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of a non-destructive testing system for the surface hardness of a steel ball provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a non-destructive testing device for the surface hardness of a steel ball provided in an embodiment of the present invention. Detailed Implementation

[0019] The specific scheme of the non-destructive testing method for the surface hardness of steel balls provided by the present invention will be described in detail below with reference to the accompanying drawings.

[0020] This invention proposes a non-destructive testing method for the surface hardness of steel balls. Please refer to [link / reference]. Figure 1 The diagram illustrates a schematic flowchart of a non-destructive testing method for the surface hardness of a steel ball according to an embodiment of the present invention. The method includes the following steps: Step S1: Obtain initial hardness detection data and corresponding spatial location information of multiple detection points on the surface of the steel ball to be tested; the initial hardness detection data includes complex impedance eigenvectors; the complex impedance eigenvectors are used to characterize the electromagnetic response characteristics of the steel ball to be tested at the detection points.

[0021] Among them, the steel balls to be tested refer to the finished wear-resistant steel balls that have completed the tempering process and are transported to the quality inspection station at the end of the production line. They are the test objects for non-destructive hardness testing and have undergone a complete heat treatment process of heating, forging, quenching, controlled cooling, and tempering.

[0022] Multiple detection points on the surface refer to the reference sampling points that are uniformly distributed on the three-dimensional surface of the steel ball based on the spherical Fibonacci equal area grid division algorithm. They can cover the entire surface from the equator to the poles and are divided into two categories: normal detection points and abnormal detection points.

[0023] Initial hardness test data refers to the complex impedance characteristic vector directly acquired and demodulated by the eddy current sensor. It is the original electromagnetic hardness characterization data without weight correction or global fusion, including the original responses of normal points and defect points.

[0024] Spatial location information refers to the radius of the steel ball cross-section corresponding to each detection point, which is obtained by the angle encoder collecting the boom deflection angle during eddy current detection and converting it with the truth table of sphere geometry mapping. It is the basis for determining the geometric coordinates of the area to which the detection point belongs.

[0025] The complex impedance eigenvector refers to the expression form of the initial hardness detection data, including the change in the equivalent resistance of the detection coil (representing the active power loss of eddy current excitation) and the change in the equivalent inductance of the detection coil (representing the reactive energy storage of the electromagnetic field). Together, they completely record the electromagnetic feedback at the detection point.

[0026] Electromagnetic response characteristics refer to the signal characteristics formed by the eddy currents induced by the alternating excitation magnetic field penetrating the surface of the steel ball, and the disturbance of the impedance and induced electromotive force of the original excitation coil by the reverse magnetic field of the eddy currents. These characteristics are determined by the metallographic structure of the steel ball surface, such as the proportion of martensite, carbides, and lattice distortion (which directly determine hardness). Electromagnetic response characteristics are the intermediate carrier for establishing a mapping relationship between electromagnetic signals and hardness.

[0027] Step S2: Divide multiple detection points into corresponding spatial rings based on spatial location information, and within each spatial ring, divide multiple detection points into standard detection groups and abnormal detection point sets based on the distribution characteristics of the initial hardness test data; the spatial rings are used to divide the surface of the steel ball to be tested into multiple analysis areas radially.

[0028] Among them, the spatial ring zone refers to the complete radius interval defined by the maximum cross-sectional radius of the steel ball at the equator and the minimum cross-sectional radius at the poles, and the independent radius sub-intervals that are divided into non-overlapping and continuously connected sub-intervals according to a fixed radius step size. Each sub-interval corresponds to a ring-shaped scanning area on the surface of the steel ball, which is used to group detection points according to the geometric position of the spherical surface.

[0029] Distribution characteristics refer to the statistical patterns of data where the signal magnitude values ​​of normal detection points within the same ring are concentrated and fluctuate only within the system noise range, while the magnitude values ​​of abnormal detection points deviate from the group and exhibit isolated and discrete distribution.

[0030] The standard test group refers to the set of normal test points where the inner mold length of a single ring falls within the deviation threshold range, the corresponding heat treatment metallographic structure is uniform, and the hardness is qualified. It represents the true hardness reference signal of the ring.

[0031] An abnormal detection point set refers to the set of all abnormal detection points within the same spherical ring where the modulus residual exceeds the deviation threshold range, or where the abnormality is caused by structural deterioration such as decarburization / quenching soft spots during heat treatment. The electromagnetic response of these abnormal points varies with the radius and has an inherent deviation from that of normal points.

[0032] Furthermore, the step of dividing multiple detection points into a standard detection group and an abnormal detection point set based on the distribution characteristics of the initial hardness test data within each spatial annulus includes: For each spatial ring, calculate the mean and standard deviation of the initial hardness test data for all test points within the spatial ring.

[0033] Among them, the mean It refers to the arithmetic mean of the initial hardness test data of all test points within a single spatial ring, representing the reference center level of the normal hardness signal in that ring region.

[0034] Standard deviation It refers to the degree of dispersion of the initial hardness test data of all test points within a single spatial ring relative to the mean, quantifying the fluctuation amplitude of normal random noise in the signal within that ring.

[0035] Based on the mean and standard deviation, the range of detection data deviation corresponding to the spatial ring zone is determined.

[0036] Among them, the deviation range of the detection data refers to the range of deviations based on... The interval defined by the criteria This represents the normal signal fluctuation range allowed by the current ring belt process, which is caused only by mechanical vibration of the equipment and circuit noise.

[0037] Test points whose initial hardness test data are within the test data deviation range are classified into standard test groups, and test points whose initial hardness test data are not within the test data deviation range are classified into abnormal test point sets.

[0038] For example, taking a single spatial ring as an independent statistical unit, the original modulus data of electromagnetic hardness of all test points in the same ring area are first collected. The mean and standard deviation of this set of data are calculated through statistical operations, and then industrially common methods are used. The normal distribution discrimination rule defines the normal fluctuation boundary of the signal; then, the initial hardness data of each detection point in the ring are compared between intervals. If the data falls within the boundary, it is classified into the standard detection group representing qualified metallographic structure. If the data exceeds the boundary, it is uniformly collected to form an abnormal detection point set characterizing heat treatment defects, thus completing the binary classification labeling of all sampling points on the spherical surface.

[0039] Step S3: Calculate the radius modulus difference between the standard detection group and the abnormal detection point set, and analyze the radius modulus difference to obtain the heterogeneity of the abnormal detection points; the radius modulus difference is used to characterize the degree of deviation between the electromagnetic response characteristics of the abnormal detection point set and the standard detection group as a function of radius; the heterogeneity of the abnormal detection points is used to characterize the deviation stability of the electromagnetic response behavior of the abnormal detection point set relative to the standard detection group during the detection process.

[0040] Among them, the radius-modulus difference refers to the absolute difference between the asynchronous evolution of radius and modulus of the normal standard detection group and the asynchronous evolution of radius and modulus of the abnormal detection point. It quantifies the difference in the evolution law of eddy current signal modulus of normal and abnormal points within the same ring as the radius of the sphere changes. The larger the value, the more significant the difference in the trend of change between the two.

[0041] The radius-modulus evolution law refers to the inherent corresponding change relationship between the continuous decrease in the radius of the spherical cross section, the nonlinear decrease in the probe cutting speed, and the synchronous attenuation of the eddy current signal complex impedance modulus when the steel ball is scanned from the equator to the poles. It is the basic characteristic curve formed by the coupling of the steel ball's geometry and electromagnetic detection.

[0042] The heterogeneity of anomaly detection points refers to a quantitative index calculated by weighting the residual of time-series fluctuations in radius modulus difference and the uniformity of abnormal data as parameters. The smaller the value, the more stable the deviation behavior of the electromagnetic response of the anomaly point with the change of radius and the stronger the concealment of the defect. The larger the value, the more chaotic the fluctuation of the abnormal signal and the easier it is to identify the defect.

[0043] Deviation stability refers to whether the radius-modulus variation deviation characteristics corresponding to the abnormal detection point remain uniform and without drastic random jumps during continuous sampling and multi-point detection. Higher stability indicates that the metallographic deterioration in the defect area is uniform and the electromagnetic abnormal signal is easily masked by the spherical geometric speed change effect.

[0044] Furthermore, the step of calculating the radius modulus difference between the standard detection group and the abnormal detection point set includes: Calculate the asynchronous degree of the first radius modulus corresponding to the standard detection group and the asynchronous degree of the second radius modulus corresponding to the abnormal detection point set.

[0045] Among them, the first radius modulus asynchrony refers to the morphological correlation coefficient calculated by Pearson correlation coefficient between the average scanning radius sequence of the annular band and the vector modulus sequence within the standard detection group. Then, through the formula The calculated value represents the degree of synchronous matching between the cross-sectional radius of the normal detection point and the modulus of the eddy current signal as the position of the sphere changes.

[0046] The second radius modulus asynchronous degree refers to the asynchronous degree calculated by the logic after each anomaly detection point constructs its own radius and modulus curve. This characterizes the degree of synchronous matching between the cross-sectional radius of the defect location and the modulus of the eddy current signal as the spherical position changes.

[0047] The radius modulus difference is obtained based on the absolute value of the difference between the first radius modulus asynchrony degree and the second radius modulus asynchrony degree.

[0048] Among them, the absolute value of the difference ( This is used to eliminate positive and negative bias interference, retaining only the deviation of the radius-modulus evolution law of the two sets of samples.

[0049] For example, by quantifying the synchronous characteristics of the radius and eddy current modulus changes of normal and abnormal detection samples respectively, and then taking the absolute value of the difference between the two asynchronous degrees to obtain the radius and modulus difference, it is possible to eliminate the uniform signal drift interference caused by the nonlinear attenuation of the geometric linear velocity of the steel ball surface, and simply separate the differences in electromagnetic response law caused by heat treatment defects, so as to provide reliable basic characteristic quantities that are not affected by the geometric structure of the sphere for subsequent processing.

[0050] It should be noted that, in the embodiments of the present invention, firstly, spatial rings are divided for all normal detection points, the average cross-sectional radius and average eddy current vector magnitude of each ring are extracted, two sets of variation sequences are constructed and normalized, and the similarity of the two sets of curve evolution is obtained using the Pearson correlation coefficient. Calculate the asynchronous degree of the first radius modulus of the standard test group. Then, for each anomaly detection point, its own radius and modulus variation curves are established. Using the same normalization and correlation coefficient conversion process, the asynchrony degree of the second radius and modulus of the anomaly detection point set is obtained. Finally, subtract the two asynchronous degrees and take the absolute value to eliminate the cancellation problem caused by positive and negative offsets, and output the radius-to-length difference. The entire process uses a unified calculation rule to process normal and abnormal samples, ensuring that the two sets of features have an equivalent benchmark for horizontal comparison.

[0051] Furthermore, the calculation steps for the asynchronous degree of the first radius modulus corresponding to the standard detection group include: Obtain the radius of the first cross-section circle and the magnitude of the first vector corresponding to each test point in the standard test group.

[0052] The radius of the first cross-section circle refers to the radius of the horizontally cut ring of the steel ball corresponding to each detection point in the standard detection group. It represents the spatial position of the probe scanning the spherical surface, with the largest radius at the equator and close to 0 at the poles.

[0053] The first vector magnitude refers to the complex impedance characteristic vector acquired by the eddy current sensor. model , represents the total voltage amplitude of the eddy current after demodulation, and is an electromagnetic characterization of the surface hardness and metallographic structure of the steel ball.

[0054] The first cross-sectional circle radius and the first vector magnitude are serialized to obtain the scan radius sequence and the vector magnitude sequence, respectively.

[0055] The serialization process refers to first calculating the mean cross-sectional radius and the mean vector magnitude of all detection points in each standard detection group, then sorting all detection group mean values ​​from largest to smallest by radius to form an ordered sequence, and simultaneously performing minimum-maximum normalization (Min-Max) to eliminate dimensional differences.

[0056] The scanning radius sequence refers to an ordered dimensionless sequence formed by arranging the average cross-sectional radii of each standard detection group from largest to smallest after serialization, reflecting the continuous variation of the spherical scanning position.

[0057] The vector magnitude sequence refers to the ordered dimensionless sequence formed by sorting the average vector magnitudes of each standard test group according to their corresponding radii after serialization processing. It reflects the electromagnetic response law of normal hardness samples as the spherical position changes.

[0058] Calculate the first morphological correlation coefficient between the scan radius sequence and the vector magnitude sequence.

[0059] Among them, the first morphological correlation coefficient refers to the quantitative value calculated using the Pearson correlation coefficient, which characterizes the overall evolution trend and morphological similarity between the scan radius change curve and the vector magnitude change curve.

[0060] Based on the correlation coefficient of the first morphology, the asynchronous degree of the first radius modulus is calculated.

[0061] For example, in an embodiment of the present invention, the cross-sectional radius and eddy current complex impedance modulus of all qualified standard test points are first extracted. The mean value within the group is calculated using a single ring as a unit to eliminate single-point random noise. Then, all ring mean values ​​are sorted by radius from the equator to the poles to complete serialization. The radius and modulus values ​​are unified to the [0,1] interval by Max-Min normalization to eliminate interference from size and signal amplitude dimensions. Subsequently, the overall trend of the two sequence curves is matched with the Pearson correlation coefficient to output the first morphological correlation coefficient characterizing the morphological similarity. Finally, the first radius and modulus asynchrony is obtained by subtracting the first morphological correlation coefficient from 1, thus completing the quantitative calibration of the synchronization degree between the normal sample position and the hardness signal change.

[0062] Furthermore, the calculation steps for the asynchronous degree of the second radius modulus corresponding to the anomaly detection point set include: Obtain the radius of the second cross-section circle and the magnitude of the second vector corresponding to each detection point in the anomaly detection point set.

[0063] The second cross-sectional radius refers to the horizontal cross-sectional radius of the abnormal detection point on the surface of the steel ball. It is different from the cross-sectional radius of the standard detection group. As a unique geometric position parameter for abnormal points, it is used to characterize the geometric position of the point from the equator / polarity of the steel ball and reflects the metal cutting line speed during probe scanning.

[0064] The second vector magnitude refers to the complex impedance characteristic vector of the anomaly detection point. Modulo calculation , represents the total voltage amplitude of the eddy current signal at the abnormal point, and is the electromagnetic response intensity corresponding to the defective metallographic structure, which is different from the vector magnitude of the normal detection group.

[0065] The radius of the second cross-section circle and the magnitude of the second vector are serialized to obtain the abnormal radius sequence and the abnormal magnitude sequence.

[0066] Here, the anomaly radius sequence refers to a one-dimensional ordered sequence consisting of the cross-sectional radii of all anomaly detection points after normalization and ordered sorting. It is used to describe the geometric distribution of abnormal points as their position on the sphere changes.

[0067] Anomaly magnitude sequence refers to an ordered sequence of anomaly point vector magnitudes that corresponds one-to-one with the anomaly radius sequence and is synchronously sorted and normalized. This describes the evolution of the electromagnetic response intensity at anomaly points as a function of the spherical position.

[0068] Calculate the second morphological correlation coefficient between the abnormal radius sequence and the abnormal modulus sequence.

[0069] The second-morphological correlation coefficient refers to the quantized value calculated using the Pearson correlation coefficient algorithm. This characterizes the similarity in the overall evolution trend and morphological trajectory of the two curves, the abnormal radius sequence and the abnormal modulus sequence, corresponding to the morphological correlation coefficient calculated by the standard detection group. This is the first-order correlation coefficient.

[0070] Based on the correlation coefficient of the second morphology, the asynchronous degree of the second radius modulus is calculated.

[0071] For example, the second radius modulus is asynchronous. It can be represented as: in, This represents the correlation coefficient of the second morphology.

[0072] In an embodiment of the present invention, firstly, each defect point within the anomaly detection point set is traversed, and the second cross-sectional circle radius (spherical geometric position parameter) and second vector magnitude (defect electromagnetic response amplitude) matching each point are extracted. Max-Min normalization is then performed on both sets of parameters to eliminate dimensional influence. Finally, the parameters are uniformly sorted according to the cross-sectional radius from largest to smallest (equatorial to polar region) to generate an ordered sequence of anomaly radii and anomaly magnitudes. Subsequently, the Pearson correlation coefficient algorithm is used to globally fit the numerical change curves of the two sequences to obtain the second morphological correlation coefficient, which characterizes the similarity of the overall change trends of the anomaly point radius and magnitude. Finally, a fixed formula is used... The second radius modulus asynchrony degree is calculated to complete the quantitative output of the geometric-electromagnetic evolution synchronization characteristics of the anomaly point itself. This result is then compared with the first radius modulus asynchrony degree of the standard detection group to obtain the radius modulus difference degree.

[0073] Furthermore, the step of analyzing the difference in radius modulus to obtain the heterogeneity of abnormal detection points includes: Obtain time-series data on the variation of radius modulus difference over time.

[0074] Among them, the detection time series data refers to the continuous time series composed of the normalized radius modulus difference corresponding to each sampling moment during the full surface scanning process, which is continuously collected at a fixed sampling interval (e.g., 0.02s). It is used to dynamically observe the real-time fluctuation of the difference between abnormal and normal data patterns as the scanning process progresses.

[0075] Based on the detection time series data, the fluctuation residual corresponding to the radius modulus difference is calculated; the fluctuation residual is used to characterize the instability of the radius modulus difference.

[0076] Among them, the fluctuation residual refers to the absolute difference between the normalized radius modulus of two adjacent sampling times. It intuitively represents the jump amplitude of the difference between the radius modulus of the two times. The larger the difference, the worse the temporal stability of the difference.

[0077] The instability of radius modulus difference refers to the overall level of fluctuation residual, which represents whether the difference in the evolution of signals between normal and abnormal detection points during continuous sampling of the entire surface is frequent and drastic. It is a temporal characteristic for judging the concealment of abnormal defects.

[0078] The uniformity of anomalous data is calculated relative to a baseline anomalous detection point. The baseline anomalous detection point is an anomalous detection point selected from the anomalous detection point set during the initial detection phase to provide data reference. The uniformity of anomalous data is used to characterize the consistency of electromagnetic response characteristics within the anomalous detection point set.

[0079] Among them, the benchmark anomaly detection point refers to the set of original anomaly detection points selected in the initial Fibonacci grid sampling stage. Its eddy current mode length sequence serves as a fixed reference standard to compare the electromagnetic response stability of newly added anomaly points.

[0080] Based on the fluctuation residuals and the uniformity of outlier data, the heterogeneity of outlier detection points is calculated.

[0081] Among them, the uniformity of abnormal data refers to the maximum topological distance between the cumulative distribution function of the baseline abnormal modulus sequence and the real-time newly added abnormal modulus sequence calculated by the KS test. The smaller the value, the more uniform the eddy current electromagnetic response of all abnormal points in the ring and the more consistent the deterioration characteristics of the defect metallographic structure.

[0082] For example, heterogeneity of anomaly detection points It can be represented as: in, This indicates that a single spatial ring band, within the current detection time period, is composed of... The total number of sets of first-order adjacent radius modulus difference fluctuation residuals generated from each time-series sampling point satisfies the following condition. , Represents the count of mutation features. The fluctuation residual represents the degree of difference between adjacent time points. Indicates the uniformity of outlier data. Represents the hyperbolic tangent function. This represents the natural exponential function. This indicates the preset sensitivity gain coefficient (e.g., 1.0).

[0083] It should be noted that, in the embodiments of the present invention, firstly, the difference in normalized radius modulus at each time point is continuously collected at a fixed sampling period to construct a complete time series sequence; secondly, the absolute difference of the difference at adjacent sampling times is calculated to obtain the fluctuation residual, and the count of abrupt change features that occur in the time series is counted; then, using the modulus sequence of the benchmark anomaly detection points selected in the initial stage as a reference, the modulus sequence of the newly added anomaly points in real time is subjected to the KS distribution test with the benchmark sequence to calculate the uniformity of the anomaly data at each time point; finally, the fluctuation residual, abrupt change count, and anomaly data uniformity are weighted and fused through exponential and hyperbolic tangent constraints to obtain the anomaly detection point heterogeneity that finally characterizes the degree of defect concealment.

[0084] For example, in a statistical time series, when adjacent fluctuation residuals When the frequency exceeds the preset jump threshold, a mutation is recorded, and the total frequency is recorded as the mutation feature count g.

[0085] Step S4: Determine the weight coefficients corresponding to the anomaly detection point set based on the heterogeneity of the anomaly detection points.

[0086] Among them, abnormal detection points refer to those that have been detected by... Threshold screening identifies spherical sampling points where the initial hardness electromagnetic modulus data exceeds the normal fluctuation range, corresponding to surface decarburization, quenching soft spots, metallographic deterioration, and substandard hardness.

[0087] The weight coefficient corresponding to the abnormal detection point set refers to the corrected weight after multiplying the initial evaluation weight and the weight adjustment coefficient. It is the actual contribution ratio of the abnormal point data to the global hardness weighted calculation. The higher the concealment of the defect, the smaller the weight value. That is, the weight coefficient is positively correlated with the heterogeneity of the abnormal detection points.

[0088] For example, the higher the heterogeneity of anomaly detection points, the larger the weight adjustment coefficient, and the higher the actual contribution weight of the anomaly points; the lower the heterogeneity, the smaller the weight, and the increasing and decreasing trends of the two are consistent.

[0089] Furthermore, the step of determining the weight coefficients corresponding to the anomaly detection point set based on the heterogeneity of the anomaly detection points includes: The weight adjustment coefficient is calculated based on the heterogeneity of anomaly detection points and the preset threshold.

[0090] The preset threshold (e.g., 0.3) refers to the pre-set critical value for activating the weight reduction penalty, serving as a boundary for distinguishing between minor anomalies that do not require weight reduction and highly hidden defects that require significant weight reduction.

[0091] The weight adjustment coefficient refers to the nonlinear correction coefficient constructed based on the sigmoid exponential function. It is solved by normalized anomaly detection point heterogeneity, threshold, and nonlinear gain coefficient. The lower the heterogeneity (the more hidden the defect), the smaller the coefficient value, which is used to attenuate the weight of the abnormal data.

[0092] For example, the weighting adjustment coefficient It can be represented as: in, This represents the natural exponential function. This represents the preset nonlinear gain function (e.g., 10.0). This represents the heterogeneity of the outlier detection point after normalization of the s-th spatial ring. This indicates the preset threshold value (e.g., 0.3).

[0093] It should be noted that, in the embodiments of the present invention, based on the historical prior detection sample library, the historical maximum heterogeneity benchmark and minimum heterogeneity benchmark are extracted; the maximum-minimum normalization algorithm is used to map the calculated anomaly detection point heterogeneity to the [0,1] interval, thus obtaining the normalized anomaly detection point heterogeneity. .

[0094] For example, the nonlinear gain coefficient refers to a preset constant (e.g., a value of 10.0) that controls the steepness of the decay of the weight adjustment coefficient and adjusts the sensitivity of the weight reduction to hidden abnormal data.

[0095] Multiply the initial evaluation weights corresponding to the anomaly detection point set by the weight adjustment coefficients to obtain the weight coefficients corresponding to the anomaly detection point set.

[0096] The initial evaluation weight corresponding to the abnormal detection point set refers to the unified basic weight assigned to all abnormal detection points in the ring zone before any correction. The default global weight is 1.0, which means that abnormal data participates in the hardness evaluation equally with normal data when no penalty is imposed.

[0097] Step S5: Correct the initial hardness test data of the abnormal detection point set based on the weight coefficient to obtain the corrected hardness test data, and then perform weighted fusion with the initial hardness test data of the standard test group to obtain the target hardness test result.

[0098] Among them, the corrected hardness test data refers to the electromagnetic characterization data after the division of abnormal test points has been weighted and penalized according to the weight of the normalized abnormal test point heterogeneity calculation of the spherical annulus to which they belong, which greatly reduces the data contribution of defect points in the evaluation.

[0099] Weighted fusion refers to the unified integration of the original data of the high-confidence standard test group and the outlier correction data after weight reduction, and the normalized weighted calculation according to the weight of each group of data to complete the integration of global surface hardness information.

[0100] The target hardness test result refers to the overall true and pure surface hardness index of the steel ball after integrating the hardness inversion values ​​of all normal points and weighted abnormal points of the global surface. It is the basis for judging the hardness of the final product after eliminating the interference of defect data.

[0101] For example, in an embodiment of the present invention, the weight coefficients corresponding to each spherical ring are first retrieved, and the original initial hardness test data of each group of abnormal detection points is multiplied with the corresponding weight to complete the correction process, generating corrected hardness test data that weakens the contribution of defects; then, the data sources are distinguished, the uncorrected initial hardness test data of the standard test group is completely retained, and the two types of data are uniformly incorporated into the global evaluation model. The normal data and the corrected abnormal data are respectively assigned corresponding confidence weights to carry out weighted fusion calculation, and finally the target hardness test result that integrates the information of all detection points on the global surface is output.

[0102] Furthermore, the step of weightedly fusing the corrected hardness test data with the initial hardness test data of the standard test group to obtain the target hardness test result includes: The initial hardness test data and the corrected hardness test data of the standard test group are input into the preset hardness inversion model to obtain the corresponding hardness inversion values.

[0103] The preset hardness inversion model is a multivariate nonlinear polynomial regression model that is calibrated and fitted using multiple sets of steel balls with known standard Rockwell hardness before the steel ball testing equipment goes online. This model is then stored in the main control computer and can be directly called up during online testing. The model uses the eddy current complex impedance characteristic vector magnitude, impedance plane phase angle, and the radius of the spherical cross-section at the testing point as input variables. It uses higher-order interaction terms of variables to remove systematic signal interference caused by probe lift-off noise and spherical scanning linear velocity attenuation, thereby eliminating the nonlinear coupling relationship between multiple physical quantities. The electromagnetic response data characterizing the metallographic structure and hardness state of the steel ball surface is quantitatively converted into a single-point Rockwell hardness value, realizing real-time mapping from electromagnetic signals to actual hardness.

[0104] Hardness inversion value refers to the Rockwell hardness value of a single point output by the hardness inversion model after inputting a single set of electromagnetic test data. It is divided into two categories: the original inversion value of normal points in the standard group and the inversion value after weighted correction of abnormal points.

[0105] The target hardness test result is obtained by weighted averaging of all hardness inversion values.

[0106] The weighted average calculation refers to the operation of using the evaluation weight corresponding to the spatial ring to which each test point belongs as the weighting coefficient, summing all the single-point hardness inversion values ​​with weights, and then dividing by the total weight. The evaluation weight of the spherical standard normal test point is fixed at 1.0. The weight of the abnormal test points in each spatial ring is reduced by solving the weight adjustment coefficient based on the heterogeneity of the normalized abnormal test points in the corresponding ring, thereby weakening the interference of the distorted hardness data of the tissue deterioration defect test points on the global evaluation results.

[0107] It should be noted that, in the embodiments of the present invention, the original standard group data and the abnormal data after weight reduction correction are distinguished and hardness inversion is performed separately. Then, all hardness inversion values ​​are fused by weighted average. This not only preserves the weight ratio of the true hardness signal of qualified test points, but also adaptively reduces the influence of distorted hardness data caused by heat treatment defects, avoids the problem of missed detection caused by defect points raising the overall hardness, and outputs an overall hardness result that fits the actual metallographic structure of the steel ball.

[0108] In summary, this invention relies on a radial spatial annular zone division strategy on the spherical surface of a steel ball. Combined with electromagnetic response data characterized by complex impedance eigenvectors, it distinguishes between the standard detection group and the abnormal detection point set. This separates defective test points caused by heat treatment decarburization and low hardness from metallographically acceptable test points, addressing the limitation of conventional algorithms in decoupling normal and abnormal detection data. The deviation of the electromagnetic response from the standard detection group to the abnormal detection point set as a function of the spherical radius is quantified by calculating the difference in radius modulus between the two sets. Then, the heterogeneity of the abnormal detection points is solved based on time-series fluctuation characteristics, reflecting the degree of concealment of defect signals. Adaptive weighting coefficients are dynamically generated based on the heterogeneity of the abnormal detection points to reduce the weighting correction of the original electromagnetic data of the abnormal detection points, weakening the interference of distorted data from deteriorated structures on the overall evaluation. Finally, the original data from the standard group and the corrected abnormal data are weighted and used to calculate the target hardness detection result. This not only fully covers the detection information of the entire steel ball surface, avoiding missed detections of local defects, but also eliminates system signal interference caused by the nonlinear attenuation of the spherical scanning linear velocity, thereby improving the non-destructive testing effect of steel ball surface hardness.

[0109] This invention proposes a non-destructive testing system for the surface hardness of steel balls. Please refer to [link / reference]. Figure 2 The diagram illustrates a structural schematic of a non-destructive testing system 200 for the surface hardness of a steel ball according to an embodiment of the present invention. The system includes: The acquisition module 201 is used to acquire the initial hardness detection data and corresponding spatial position information of multiple detection points on the surface of the steel ball to be tested; the initial hardness detection data includes a complex impedance feature vector; the complex impedance feature vector is used to characterize the electromagnetic response characteristics of the steel ball to be tested at the detection point. The partitioning module 202 is used to divide multiple detection points into corresponding spatial rings based on spatial location information, and within each spatial ring, based on the distribution characteristics of the initial hardness test data, divide the multiple detection points into a standard detection group and an abnormal detection point set; the spatial rings are used to divide the surface of the steel ball to be tested into multiple analysis areas along the radial direction; Analysis module 203 is used to calculate the radius modulus difference between the standard detection group and the abnormal detection point set, and to analyze the radius modulus difference to obtain the heterogeneity of the abnormal detection points; the radius modulus difference is used to characterize the degree of deviation between the electromagnetic response characteristics of the abnormal detection point set and the standard detection group as a function of radius; the heterogeneity of the abnormal detection points is used to characterize the deviation stability of the electromagnetic response behavior of the abnormal detection point set relative to the standard detection group during the detection process. The determination module 204 is used to determine the weight coefficients corresponding to the set of anomaly detection points based on the heterogeneity of the anomaly detection points. The correction module 205 is used to correct the initial hardness test data of the abnormal detection point set based on the weight coefficient, so as to obtain the corrected hardness test data. The fusion module 206 is used to perform weighted fusion of the corrected hardness test data and the initial hardness test data of the standard test group to obtain the target hardness test result.

[0110] It should be noted that the system provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. In addition, the non-destructive testing system for the surface hardness of steel balls and the non-destructive testing method for the surface hardness of steel balls provided in the above embodiments belong to the same concept, and the specific implementation process is detailed in the method embodiments, which will not be repeated here.

[0111] This invention also provides a non-destructive testing device for the surface hardness of steel balls. Please refer to [link / reference]. Figure 3 The non-destructive testing equipment 300 for steel ball surface hardness may include a non-destructive testing system 200 for steel ball surface hardness, and the non-destructive testing equipment 300 for steel ball surface hardness can achieve... Figure 1 Any steps in the corresponding method embodiments and the achievement of the same beneficial effects will not be repeated here.

Claims

1. A non-destructive testing method for the surface hardness of a steel ball, characterized in that, The method includes: Acquire initial hardness detection data and corresponding spatial location information of multiple detection points on the surface of the steel ball to be tested; the initial hardness detection data includes a complex impedance feature vector; the complex impedance feature vector is used to characterize the electromagnetic response characteristics of the steel ball to be tested at the detection point. Based on the spatial location information, the multiple detection points are divided into corresponding spatial rings, and within each spatial ring, based on the distribution characteristics of the initial hardness test data, the multiple detection points are divided into a standard detection group and an abnormal detection point set; the spatial rings are used to divide the surface of the steel ball to be tested into multiple analysis regions radially; The radius modulus difference between the standard detection group and the abnormal detection point set is calculated, and the radius modulus difference is analyzed to obtain the heterogeneity of the abnormal detection points. The radius modulus difference is used to characterize the degree of deviation between the electromagnetic response characteristics of the abnormal detection point set and the standard detection group as a function of radius. The heterogeneity of the abnormal detection points is used to characterize the deviation stability of the electromagnetic response behavior of the abnormal detection point set relative to the standard detection group during the detection process. Based on the heterogeneity of the anomaly detection points, determine the weight coefficients corresponding to the set of anomaly detection points; The initial hardness detection data of the abnormal detection point set is corrected based on the weighting coefficient to obtain corrected hardness detection data. The corrected hardness detection data is then weighted and fused with the initial hardness detection data of the standard detection group to obtain the target hardness detection result.

2. The non-destructive testing method for the surface hardness of a steel ball according to claim 1, characterized in that, The step of dividing the multiple test points into a standard test group and an abnormal test point set based on the distribution characteristics of the initial hardness test data within each spatial annulus includes: For each of the aforementioned spatial rings, calculate the mean and standard deviation of the initial hardness test data for all test points within the spatial ring; Based on the mean and the standard deviation, the detection data deviation range corresponding to the spatial ring zone is determined; The test points whose initial hardness test data is within the range of the test data deviation are classified into the standard test group, and the test points whose initial hardness test data is not within the range of the test data deviation are classified into the abnormal test point set.

3. The non-destructive testing method for the surface hardness of a steel ball according to claim 1, characterized in that, The step of calculating the radius modulus difference between the standard detection group and the abnormal detection point set includes: Calculate the first radius modulus asynchrony degree corresponding to the standard detection group and the second radius modulus asynchrony degree corresponding to the abnormal detection point set, respectively; The radius modulus difference is obtained based on the absolute value of the difference between the first radius modulus asynchrony degree and the second radius modulus asynchrony degree.

4. The non-destructive testing method for the surface hardness of a steel ball according to claim 1, characterized in that, The step of analyzing the difference in radius modulus to obtain the heterogeneity of abnormal detection points includes: Acquire the detection time-series data of the radius modulus difference over time; Based on the detection time series data, the fluctuation residual corresponding to the radius modulus difference is calculated; the fluctuation residual is used to characterize the instability of the radius modulus difference. The uniformity of abnormal data in the set of abnormal detection points relative to the reference abnormal detection points is calculated; the reference abnormal detection points are abnormal detection points selected from the set of abnormal detection points in the initial detection stage to provide data reference; the uniformity of abnormal data is used to characterize the consistency of electromagnetic response characteristics within the set of abnormal detection points. Based on the fluctuation residual and the uniformity of the abnormal data, the heterogeneity of the abnormal detection points is calculated.

5. The non-destructive testing method for the surface hardness of a steel ball according to claim 1, characterized in that, The step of determining the weight coefficients corresponding to the set of anomaly detection points based on the heterogeneity of the anomaly detection points includes: Based on the heterogeneity of the anomaly detection points and the preset threshold, the weight adjustment coefficient is calculated. The initial evaluation weight corresponding to the anomaly detection point set is multiplied by the weight adjustment coefficient to obtain the weight coefficient corresponding to the anomaly detection point set.

6. The non-destructive testing method for the surface hardness of a steel ball according to claim 1, characterized in that, The step of weightedly fusing the corrected hardness test data with the initial hardness test data of the standard test group to obtain the target hardness test result includes: The initial hardness test data and the corrected hardness test data of the standard test group are respectively input into the preset hardness inversion model to obtain the corresponding hardness inversion values. The target hardness test result is obtained by performing a weighted average calculation on all the hardness inversion values.

7. The non-destructive testing method for the surface hardness of a steel ball according to claim 3, characterized in that, The calculation steps for the first radius modulus asynchrony degree corresponding to the standard detection group include: Obtain the radius of the first cross-section circle and the magnitude of the first vector corresponding to each detection point in the standard detection group; The first cross-sectional circle radius and the first vector magnitude are serialized to obtain the scan radius sequence and the vector magnitude sequence, respectively. Calculate the first morphological correlation coefficient between the scan radius sequence and the vector magnitude sequence; Based on the first morphological correlation coefficient, the first radius modulus asynchrony degree is calculated.

8. The non-destructive testing method for the surface hardness of a steel ball according to claim 3, characterized in that, The calculation steps for the asynchronous degree of the second radius modulus corresponding to the anomaly detection point set include: Obtain the radius of the second cross-section circle and the magnitude of the second vector corresponding to each detection point in the anomaly detection point set; The radius of the second cross-section circle and the magnitude of the second vector are serialized to obtain the abnormal radius sequence and the abnormal magnitude sequence, respectively. Calculate the second morphological correlation coefficient between the abnormal radius sequence and the abnormal modulus sequence; Based on the second morphological correlation coefficient, the second radius modulus asynchrony is calculated.

9. A non-destructive testing system for the surface hardness of a steel ball, characterized in that, The system includes: The acquisition module is used to acquire initial hardness detection data and corresponding spatial position information of multiple detection points on the surface of the steel ball to be tested; the initial hardness detection data includes a complex impedance feature vector; the complex impedance feature vector is used to characterize the electromagnetic response characteristics of the steel ball to be tested at the detection point. The partitioning module is used to divide the multiple detection points into corresponding spatial rings according to the spatial location information, and within each spatial ring, based on the distribution characteristics of the initial hardness test data, divide the multiple detection points into a standard detection group and an abnormal detection point set; the spatial rings are used to divide the surface of the steel ball to be tested into multiple analysis regions radially; The analysis module is used to calculate the radius modulus difference between the standard detection group and the abnormal detection point set, and to analyze the radius modulus difference to obtain the heterogeneity of the abnormal detection points. The radius modulus difference is used to characterize the degree of deviation between the electromagnetic response characteristics of the abnormal detection point set and the standard detection group as a function of radius. The heterogeneity of the abnormal detection points is used to characterize the deviation stability of the electromagnetic response behavior of the abnormal detection point set relative to the standard detection group during the detection process. The determination module is used to determine the weight coefficients corresponding to the set of anomaly detection points based on the heterogeneity of the anomaly detection points. The correction module is used to correct the initial hardness detection data of the abnormal detection point set based on the weight coefficient, so as to obtain the corrected hardness detection data. The fusion module is used to perform weighted fusion of the corrected hardness test data and the initial hardness test data of the standard test group to obtain the target hardness test result.

10. A non-destructive testing device for the surface hardness of steel balls, characterized in that, The non-destructive testing equipment for the surface hardness of steel balls includes the non-destructive testing system for the surface hardness of steel balls as described in claim 9.

Images