Resin grinding wheel uniformity detection method and system based on AI model

Abstract

The application discloses a kind of uniformity detection method and system of resin grinding wheel based on AI model.The method comprises: synchronously collecting the visible light gray feature information and near-infrared light intensity feature information of the resin grinding wheel to be measured in rotating state;Space coordinate alignment is carried out to construct multispectral mapping field;Grinding wheel granularity parameter is identified using AI model and the size of sampling window is matched accordingly, combined with two waveband feature information and local gray mean value, local effective abrasive concentration projection value is calculated;By extracting the discrete degree parameter of full circumferential projection value and comparing with the preset threshold value, the uniformity detection conclusion is output.The application utilizes the penetration characteristics of near-infrared light combined with the surface feature mapping of visible light, eliminates the shielding effect of "floating pulp skin" on the surface of grinding wheel and environmental light interference through nonlinear correlation operator, realizes non-contact, quantitative and accurate detection of the internal abrasive distribution of grinding wheel, effectively improves the use safety and quality consistency of high-speed grinding wheel.

Description

Technical Field

[0001] This invention relates to the field of abrasive tool testing technology, specifically to a method and system for non-contact quantitative detection of abrasive distribution and resin matrix uniformity inside resin grinding wheels by combining multispectral visual data with deep learning models. Background Technology

[0002] As a core abrasive tool in precision machining, the performance of resin grinding wheels is directly affected by the uniformity of abrasive particle distribution and the density of the resin matrix. Under high-precision and high-speed grinding conditions, any inhomogeneity in the internal microstructure of the grinding wheel, such as abrasive segregation, agglomeration, or localized porosity, can lead to dynamic imbalance, grinding force fluctuations, and even cause the grinding wheel to burst during high-speed rotation, seriously threatening equipment and personnel safety. Therefore, conducting comprehensive uniformity testing on resin grinding wheels is a crucial step in ensuring their quality and safety.

[0003] Currently, the main testing methods for the uniformity of resin grinding wheels in the industry include the following categories: Firstly, there is manual visual inspection or inspection using ordinary industrial cameras. These methods mainly rely on illuminating the surface of the grinding wheel with visible light and judging uniformity by identifying differences in surface color or texture. However, during the hot-pressing and curing process of resin grinding wheels, due to surface tension and interfacial effects, the resin liquid often forms an extremely thin and dense "slurry skin" on the mold contact surface. This slurry skin produces strong visible light reflection, thus obscuring the actual distribution of abrasive particles located beneath it, leading to the illusion that "the surface looks uniform, but there are actually defects inside," resulting in a high rate of false positives and false negatives.

[0004] Secondly, the hardness tester sampling method. This method infers density uniformity by testing the hardness at several points on the grinding wheel surface. However, this method is a point-based inspection and cannot achieve full or subsurface coverage. Furthermore, the testing process is often slightly destructive, making it difficult to achieve fully automated online inspection and providing a digital distribution map for the entire grinding wheel.

[0005] Third, the dynamic balancing method. This method uses a dynamic balancing machine to measure the deviation of the center of mass of the grinding wheel during rotation. Although this method can reflect the overall mass distribution, it is a macroscopic physical characteristic evaluation and cannot locate localized small abrasive agglomerations or sparse defects. Furthermore, for thin-blade grinding wheels, the dynamic balancing data is difficult to pinpoint to specific radial or circumferential positions, lacking guidance for improving the production process.

[0006] Fourth, existing multispectral detection methods based on addition and subtraction normalization logic, such as NDVI, are not suitable for resin grinding wheel inspection. The unique surface "slurry skin" of resin grinding wheels creates strong nonlinear interference. As the resin layer on the grinding wheel surface thickens, the visible light reflection intensity increases, but the near-infrared light attenuates due to the longer penetration path, showing a downward trend. If the subtractive coupling logic of NDVI is used, the obtained value will have a significant negative shift due to the "increase and decrease" of the two signal bands. This signal offsetting effect makes it impossible for the system to effectively distinguish whether the change in value is caused by "simple resin layer thickening" or by "sparse abrasive density underneath," thus causing serious ambiguity in defect judgment in logic.

[0007] In summary, the existing technology lacks a non-contact, quantitative, and environmentally robust technology for detecting the uniformity of resin grinding wheels that can penetrate the surface resin masking layer. Summary of the Invention

[0008] To achieve the above-mentioned objectives, this invention provides a method for detecting the uniformity of resin grinding wheels based on an AI model, characterized by comprising: Step S1: Simultaneously acquire visible light grayscale characteristics and near-infrared light intensity characteristics of the resin grinding wheel under test while it is rotating; Step S2: Align the visible light grayscale feature information with the near-infrared light intensity feature information in spatial coordinates to construct a multispectral mapping field that reflects the physical properties of the surface and subsurface of the resin grinding wheel under test. Step S3: Based on the AI ​​model, identify the particle size parameters of the resin grinding wheel to be tested, and according to the sampling window size determined by the particle size parameters, perform nonlinear correlation calculation by combining the visible light grayscale feature information, the near-infrared light intensity feature information and the grayscale mean value within the sampling window to obtain the local effective abrasive concentration projection value. Step S4: Extract the local effective abrasive concentration projection value of each coordinate point in the full circumference, calculate the dispersion parameter of the local effective abrasive concentration projection value in the preset area, and output the uniformity detection conclusion of the resin grinding wheel to be tested based on the comparison result of the dispersion parameter and the preset threshold.

[0009] Further, the calculation of the local effective abrasive concentration projection value in step S3 specifically includes: multiplying the near-infrared light intensity feature value and the visible light grayscale feature value at a specific coordinate point to obtain a product term, and dividing the product term by the square of the visible light grayscale mean value within the sampling window to obtain the local effective abrasive concentration projection value.

[0010] Furthermore, the calculation of the dispersion parameter of the local effective abrasive concentration projection value within the preset area in step S4 specifically includes: calculating the variance of the local effective abrasive concentration projection value corresponding to all coordinate points in the full circumference.

[0011] Furthermore, the step S4 of outputting the uniformity detection conclusion of the resin grinding wheel under test based on the comparison result of the dispersion parameter and the preset threshold specifically includes: when the variance of the local effective abrasive concentration projection value in the whole circumference is greater than the preset threshold, it is determined that the resin grinding wheel under test has an internal abrasive uneven distribution defect.

[0012] To achieve the above-mentioned objectives, this invention also provides a uniformity detection system for resin grinding wheels based on an AI model, characterized in that it applies the uniformity detection method for resin grinding wheels based on an AI model as described above, comprising: The feature information acquisition module simultaneously acquires the visible light grayscale feature information and near-infrared light intensity feature information of the resin grinding wheel under test in a rotating state; The spatial coordinate alignment module aligns the visible light grayscale feature information with the near-infrared light intensity feature information in spatial coordinates to construct a multispectral mapping field that reflects the physical properties of the surface and subsurface of the resin grinding wheel under test. The projection value calculation module identifies the particle size parameters of the resin grinding wheel under test based on the AI ​​model, and performs nonlinear correlation calculation based on the sampling window size determined by the particle size parameters, combined with the visible light grayscale feature information, the near-infrared light intensity feature information, and the grayscale mean value within the sampling window, to obtain the local effective abrasive concentration projection value. The uniformity detection module extracts the local effective abrasive concentration projection values ​​at each coordinate point along the entire circumference, calculates the dispersion parameter of the local effective abrasive concentration projection values ​​within a preset area, and outputs the uniformity detection conclusion of the resin grinding wheel under test based on the comparison result of the dispersion parameter and a preset threshold.

[0013] To achieve the above-mentioned objectives, the present invention also provides a computer-readable storage medium, characterized in that it stores a computer program, which, when executed by a processor, implements a method for detecting the uniformity of resin grinding wheels based on an AI model as described in any of the preceding claims. The uniformity detection method and system for resin grinding wheels based on an AI model, as described in this invention, can achieve the following significant technical effects through the implementation of the above technical solution: First, this invention overcomes the physical obscuring limitation of the "slurry skin" on the surface of resin grinding wheels, achieving non-contact and precise detection of deep components. Compared to the limitations of traditional visible light detection, which can only focus on surface color, this solution cleverly solves the signal "cancellation" and obscuring problems caused by surface resin thickening by introducing the near-infrared (NIR) band and combining it with an original nonlinear coupling operator. This algorithm can use visible light characteristics to compensate for infrared transmission signals in real time, effectively eliminating interference from light fluctuations and resin color differences. It also amplifies the abrasive agglomeration or sparse features hidden beneath the subsurface from a physical perspective, greatly reducing the missed detection rate of latent defects and achieving a technological leap from "appearance inspection" to "internal density field reconstruction".

[0014] Secondly, this invention significantly improves the intelligence level of production inspection and the safety of high-speed operation through an AI-driven adaptive precision control and quantitative evaluation system. The system utilizes a lightweight convolutional neural network to automatically sense the grinding wheel grit size and dynamically adjust the sampling window size of the detection operator, ensuring optimal resolution and robustness across all production lines without frequent manual parameter intervention. This digital quantitative evaluation method transforms previously vague subjective judgments into scientific determinations of dispersion parameters (variance), accurately capturing microscopic component deviations where dynamic balance is acceptable but internal local stress concentration risks exist, thus preventing potential breakage accidents of high-speed grinding wheels during grinding. Furthermore, this non-contact, full-volume online inspection method significantly reduces inspection costs and time while ensuring extremely high detection depth, providing reliable technical assurance for the quality consistency of industrial-grade grinding tools. Attached Figure Description

[0015] Figure 1 This is a schematic flowchart of a method for detecting the uniformity of resin grinding wheels based on an AI model, as described in this invention. Figure 2 This is a schematic diagram of the structure of a resin grinding wheel uniformity detection system based on an AI model, as described in this invention. Figure 3 This is a schematic diagram of the electronic terminal structure described in this invention. Detailed Implementation

[0016] The present invention will now be described in detail with reference to the specific embodiments shown in the accompanying drawings. However, these embodiments do not limit the present invention, and any structural, methodological, or functional modifications made by those skilled in the art based on these embodiments are included within the scope of protection of the present invention.

[0017] If the present invention involves orientation (e.g., up, down, left, right, front, back, outside, inside, etc.) when described, then the orientations involved need to be defined.

[0018] The scope of the embodiments described herein includes the entire scope of the claims and all available equivalents thereof. Throughout this document, the terms “first,” “second,” etc., are used only to distinguish one element from another without requiring or implying any actual relationship or order between the elements. Indeed, a first element can also be referred to as a second element, and vice versa. Furthermore, the terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a structure, apparatus, or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a structure, apparatus, or device. Without further limitations, an element defined by the phrase “comprising one…” does not exclude the presence of other identical elements in the structure, apparatus, or device that includes said element. The various embodiments described herein are presented in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably.

[0019] The terms "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer" used in this document to indicate orientation or positional relationships are based on the orientation or positional relationships shown in the accompanying drawings and are used only for the convenience of describing this document and simplifying the description. They 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, and therefore should not be construed as limiting the invention. In the description herein, unless otherwise specified and limited, the terms "installed," "connected," and "linked" should be interpreted broadly. For example, they can refer to mechanical or electrical connections, or internal connections between two elements, or direct connections or indirect connections through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms according to the specific circumstances.

[0020] The following is in conjunction with the appendix Figures 1 to 3 This application provides a detailed description of an AI model-based method and system for detecting the uniformity of resin grinding wheels.

[0021] The method provided in this embodiment aims to solve the problem of inaccurate abrasive distribution detection caused by surface "slurry skin" in resin grinding wheel production. By introducing multispectral visual fusion logic, the grinding wheel surface is regarded as a "quasi-three-dimensional probability field" with optical transmission differences. By utilizing the differences in sensitivity of different wavelengths of light to resin and abrasive, the deep non-uniformity can be quantitatively characterized.

[0022] Step S1: Simultaneously acquire visible light grayscale characteristics and near-infrared light intensity characteristics of the resin grinding wheel under test in a rotating state.

[0023] In this step, the resin grinding wheel to be tested is mounted on a rotating spindle. When the grinding wheel rotates at a preset speed, the dual-band acquisition module located on the outer periphery of the grinding wheel is simultaneously triggered. Specifically, a visible light industrial camera captures the reflection characteristics of the grinding wheel surface, obtaining visible light grayscale feature information. Near-infrared cameras utilize the strong penetrating power of near-infrared light through resin to capture the light-blocking and scattering characteristics of abrasive particles hidden beneath the resin layer, thereby obtaining near-infrared light intensity characteristic information. .

[0024] The reason for collecting both types of information simultaneously is that visible light alone can only reflect the viscosity of the surface resin and cannot observe the internal abrasive; while near-infrared light, although penetrating, has an intensity modulated by the thickness of the surface resin. Only by obtaining synchronized data from both can a logical premise be provided for subsequently eliminating the shading effect.

[0025] Step S2: Align the visible light grayscale feature information with the near-infrared light intensity feature information in spatial coordinates to construct a multispectral mapping field that reflects the physical properties of the surface and subsurface of the resin grinding wheel under test.

[0026] Because the two cameras have different physical installation positions, the system performs affine transformation on the acquired images using a preset calibration plate to ensure pixel accuracy. The two spectral bands correspond to the same physical coordinates. The multispectral mapping field constructed through coordinate alignment endows each spatial coordinate point with dual physical properties: surface reflectance properties reflected by visible light, mapping the reflectance intensity and smoothness of the surface resin; and internal transmission properties reflected by near-infrared light, mapping the blocking and scattering characteristics of abrasive particles after penetrating the resin layer. Technically, this mapping field elevates traditional two-dimensional visual information to a higher dimension, constructing a "quasi-three-dimensional probability field" data foundation capable of characterizing the shallow surface structure of the grinding wheel.

[0027] Specifically, the synchronous acquisition module described in step S1 converts the optical reflection and transmission signals of the grinding wheel surface under test into electrical signals through the photosensitive element of an industrial camera, and then forms a two-dimensional digital image matrix through analog-to-digital conversion (A / D conversion). In the obtained digital image matrix, each pixel unit is assigned a unique spatial index coordinate according to its arrangement position in the photosensitive array. .

[0028] This coordinate point The quantized value stored at that location corresponds to the visible light grayscale feature value at that position. or near-infrared light intensity characteristic value Subsequently, through the calibration and affine transformation described in step S2, the visible light image matrix and the near-infrared image matrix are mapped to a unified physical coordinate system, thereby enabling the mapping of any physical position on the grinding wheel surface. The accurate extraction and correlation of multispectral feature values ​​provide point-to-point data input for subsequent nonlinear projection calculations.

[0029] Step S3: Based on the AI ​​model, identify the particle size parameters of the resin grinding wheel to be tested, and according to the sampling window size determined by the particle size parameters, perform nonlinear correlation calculations by combining the visible light grayscale feature information, the near-infrared light intensity feature information, and the grayscale mean value within the sampling window to obtain the local effective abrasive concentration projection value.

[0030] AI models (such as pre-trained lightweight convolutional neural networks) automatically determine the abrasive grit number (such as 60 coarse abrasive or 120 fine abrasive) by recognizing the texture features of the grinding wheel surface.

[0031] Specifically, the system pre-trains a lightweight convolutional neural network (such as MobileNetV3 or ShuffleNetV2). When the visible light grayscale image acquired in step S1 is input into the model, the convolutional layers extract spatial frequency features from the image through filters of different scales. This is because grinding wheels of different grit sizes have drastically different physical textures. Coarse-grained grinding wheels (such as No. 60): The image shows large gray-scale abrupt changes, clear particle boundaries and low spatial frequency, and the convolution kernel will capture edge features with a large radius.

[0032] Fine-grained grinding wheels (such as 120 grit): The image texture is characterized by a dense dotted distribution with a high spatial frequency, while the convolution kernel activates more high-frequency micro-texture features.

[0033] The fully connected layer at the end of the model maps these abstract features to a predefined grain size label space, thereby outputting the current grinding wheel's grain size number in real time. This grain size number is then used as an index to automatically retrieve the corresponding sampling window size M×N from a predefined parameter mapping table, thus providing the projection value for the subsequent local effective abrasive concentration. It provides a scale basis for precise calculations.

[0034] It should be noted that the above-described method of using lightweight convolutional neural networks for granular recognition is merely a preferred embodiment of the present invention, intended to demonstrate how the system achieves adaptive parameter configuration. Those skilled in the art should understand that other techniques capable of texture feature classification, such as Local Binary Pattern (LBP), Gray-Level Co-occurrence Matrix (GLCM) based on traditional image processing algorithms, or other types of deep learning architectures (such as VisionTransformer or simple multilayer perceptrons), are also feasible alternatives.

[0035] Therefore, this invention does not limit the specific algorithm model for identifying granularity numbers. The core lies in the technical logic of "dynamically triggering the reconstruction of the detection window size by automatically identifying granularity features." The above examples should not constitute a limitation on the scope of protection of this invention, and will not be elaborated further here.

[0036] Based on the granularity number, the system automatically matches the optimal sampling window size from the lookup table. .

[0037] Next, the core mathematical formula is used to calculate the projected value of the local effective abrasive concentration. :

[0038] In this formula: : Reflects the surface resin reflection intensity at the coordinate point.

[0039] This reflects the degree of abrasive obstruction after penetrating the resin layer at the coordinate point.

[0040] The denominator is the average visible light gray value within the local sampling window, which physically represents the macroscopic average resin background environment of that area.

[0041] Because the "slurry skin" on the surface of the resin grinding wheel absorbs and reflects light, and because abrasive particles have different scattering cross-sections for different wavelengths of light, the raw signal collected is the result of multiple intertwined physical variables. This application adopts a strategy of "first coupling enhancement, then reference decoupling": Molecules use visible light grayscale values Near-infrared light intensity value The product of . In a physical sense, It maps the apparent thickness and smoothness of the resin matrix in a local area. This maps the degree of abrasive particle aggregation after penetrating the resin layer. Through a series of multiplication operations, a "weighted adaptive mechanism" can be established: when the resin layer is thicker... When the signal attenuates, the stronger The reflected signal compensates for the molecules; conversely, when the resin layer is thin, The dominant position of the abrasive particles is enhanced. This coupling method ensures that the characteristic signals of the abrasive particles can be effectively captured under different resin coating thicknesses.

[0042] Eigenvalues ​​of coordinate points ( The numbers, as the raw input stream of the formula, represent spatial coordinates. Microscopic optical response at a point. Its physical significance lies in capturing "point-to-point" material mutations, which is the basis for discovering local agglomeration (such as near-infrared anomalies caused by abrasive accumulation) or local sparsity (such as visible light anomalies caused by resin enrichment).

[0043] Local sampling window ( The window is dynamically set by the AI ​​model based on the grinding wheel grit size, and its logic lies in defining the observation scale of "uniformity". For coarse-grained grinding wheels, the window is set large to filter out noise generated by individual large particles; for fine-grained grinding wheels, the window is set small to ensure detection resolution. This enables this application to have adaptive generalization ability for grinding wheels of different specifications.

[0044] Through the above logic, this application transforms complex, surface-affected visual signals into a linearized projection index. When the interior of the grinding wheel is absolutely homogeneous, the local signal changes synchronously with the background mean. It approaches a stable constant; when internal inhomogeneities occur (such as abrasive agglomeration), the abnormal gain of the near-infrared signal will break the balance constraint of the denominator, leading to... Dramatic fluctuations occurred. This design logic bypasses the physical limitation of "not being able to directly observe the interior" and achieves accurate quantitative detection of latent uniformity defects through cross-verification of multispectral information.

[0045] Step S4: Extract the local effective abrasive concentration projection value of each coordinate point in the full circumference, calculate the dispersion parameter of the local effective abrasive concentration projection value in the preset area, and output the uniformity detection conclusion of the resin grinding wheel to be tested based on the comparison result of the dispersion parameter and the preset threshold.

[0046] The system processes all the output of one revolution of the grinding wheel. The data is statistically analyzed, and its standard deviation or variance is calculated as a parameter of dispersion. If the variance exceeds a preset threshold, it indicates that there are local abrasive agglomerates or sparse areas inside the grinding wheel. This "hidden non-uniformity" will generate unbalanced forces under high-speed operation, and the system will immediately issue a defect warning and mark the location.

[0047] The preset threshold can be a quantitative judgment benchmark established based on statistical distribution patterns and the physical characteristics of standard samples. Its specific setting logic and acquisition process are described below: First, a set of grinding wheels, verified by physical destructive testing (such as slice hardness testing or industrial CT scanning) to have extremely uniform component distribution, were selected as standard samples. Under the same testing environment as the grinding wheels to be tested, the standard samples were scanned circumferentially using the method described in this application to obtain the local effective abrasive concentration projection values ​​at different coordinate points. .

[0048] Based on the dataset obtained from standard samples, the mean variance benchmark of its projected values ​​is calculated. In terms of physical logic, This represents the minimum discreteness achievable under ideal manufacturing processes, constrained by the random arrangement of abrasive grains. According to industrial quality control Based on principles or specific accuracy requirements, set a preset threshold. Its computational logic can be expressed as follows:

[0049] in, This is a preset amplification factor, and its value range is usually determined according to the grinding safety level. For example, for high-speed (linear velocity greater than 80 m / s) resin grinding wheels, in order to avoid the discrete stress caused by local agglomeration, The value is usually taken as a small safety margin (e.g., 1.2 to 1.5); while for ordinary low-speed grinding wheels, The value can be appropriately relaxed.

[0050] It should be emphasized that, due to the different natural characteristic fluctuations of grinding wheels with different grit sizes (such as 60 and 120), the preset threshold of this application... It will be associated with the granularity parameters identified by AI in step S3. The system has a pre-stored "granularity-threshold mapping table". When AI identifies a specific granularity, the system will automatically call the standard deviation benchmark corresponding to that granularity. This dynamic setting logic ensures the fairness and accuracy of the threshold determination and avoids false alarms caused by the switching of grinding wheel specifications.

[0051] The method described above for determining the threshold using the variance ratio of standard samples is merely a preferred embodiment of the present invention. Those skilled in the art should understand that other statistical indicators (such as coefficient of variation, range distribution, etc.) can also be used as evaluation parameters for the degree of dispersion, and corresponding judgment thresholds can be set accordingly. Such parameter fine-tuning based on known statistical principles is a routine operation that can be easily implemented by those skilled in the art according to actual production accuracy requirements, and does not depart from the logical protection scope of the present invention.

[0052] Example 1: High-precision consistency verification under standard production conditions This example focuses on a batch of ceramic-reinforced resin grinding wheels with specification "WA60#". This batch of grinding wheels exhibits a uniform dark gray appearance, with microscopic compositional differences difficult to discern with the naked eye. The execution process is as follows: AI-powered intelligent identification and parameter optimization: After system startup, the AI ​​model accurately identifies the abrasive particle size as 60# by analyzing the texture features of image slices during the initial rotation phase. Based on this, the system automatically retrieves a sampling window adapted for large-particle abrasives from the algorithm library. Pixels. The physical logic behind this is that coarse-grained abrasive particles have a larger projected area, and increasing the window helps filter out random noise caused by individual particles, thereby obtaining the true statistical concentration.

[0053] Reference field establishment: after scanning to coordinate points At that time, the sensor measured the visible light gray value. This reflects normal resin-coated reflection; near-infrared intensity values ​​were measured simultaneously. This demonstrates the normal infrared blocking characteristics of the abrasive. At this point, the system calculates the average background value within this local window. .

[0054] Precise quantification: Substitute the data into the formula for calculation:

[0055] Full scan and conclusion: System completed. After scanning, the generated The distribution field shows that the values ​​are stable across the entire circumference. Within the interval, the variance calculation result is as follows: It is far below the preset quality alarm threshold. .

[0056] Conclusion: The system determined that the internal composition of the grinding wheel is extremely uniform, and there is no risk of abrasive segregation. It can proceed directly to the next fine balancing process.

[0057] Example 2: Identification of "Hidden Agglomeration" under Critical Defect Conditions This example focuses on a grinding wheel exhibiting a "resin segregation" anomaly during the molding process. Due to uneven mold pressure, a dense, high-gloss resin film has formed on the wheel's surface. Under traditional visual inspection, this grinding wheel, with its bright and consistent surface color, is often mistakenly judged as a "superior product." The execution process is as follows: Optical interference sensing: Scan to coordinate point At that time, the visible light camera, due to capturing the strong reflection from the bright resin skin, returns a grayscale value. (Much higher than normal levels).

[0058] Penetrating Depth Detection: Simultaneously, the near-infrared camera, by penetrating the resin layer, captured abrasive agglomerates formed beneath due to gravity settling. Because these agglomerates exert a strong diffuse reflection and blocking coupling effect on the infrared signal, the measured depth... (Approaching saturation value).

[0059] Correlation operator operation: The system identifies the local background mean of the region. Perform nonlinear correlation calculations based on the formula:

[0060] Logical Judgment and Risk Warning: Although from a single visible light perspective, the difference between this point and the surrounding background (mean 205) is not significant, due to... Strong abnormal intervention led to A sudden surge. In the total circumferential statistics, the dispersion parameter (variance) at this point increases dramatically. This triggered the system threshold (0.04).

[0061] Execution conclusion: The system automatically adjusts the side of the grinding wheel. A red warning box was marked at the location, and a message was displayed: "Hidden component agglomeration detected, projected concentration exceeds the standard by 20%." This conclusion successfully avoided missed detections caused by surface "slurry skin" and prevented wheel breakage accidents caused by uneven local stress during high-speed grinding.

[0062] Figure 2 The detection system shown: Feature information acquisition module: Built-in visible light and near-infrared sensors, execute step S1.

[0063] Spatial coordinate alignment module: Executes the image registration algorithm and completes step S2.

[0064] Projection value calculation module: It has a built-in AI model and the above logical formula, and executes step S3.

[0065] Uniformity detection module: Performs statistical analysis and threshold determination, outputs results, and executes step S4.

[0066] Figure 3 The electronic terminal shown includes a processor, a memory, and a communication bus. The memory stores computer program instructions that implement the above-described method. When the processor calls and executes these instructions, it controls the hardware to acquire multispectral data and follows the above-described... The operator logic performs real-time calculations to achieve digital automatic detection of the uniformity of the resin grinding wheel.

[0067] The storage medium provided in this embodiment stores an encrypted binary program. After the program is read by a computer, it can drive the vision terminal on the industrial inspection line to execute all the above steps, ensuring the consistency and traceability of the inspection logic.

[0068] This invention can be an apparatus, method, and / or computer program product. A computer program product may include a readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of the invention.

[0069] Storage media can be tangible devices that hold and store instructions for use by instruction execution devices. Storage media can include, for example, electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination thereof.

[0070] It should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This way of describing the specification is only for clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

[0071] The detailed descriptions listed above are merely specific descriptions of feasible embodiments of the present invention, and are not intended to limit the scope of protection of the present invention. All equivalent embodiments or modifications made without departing from the spirit of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for detecting the uniformity of resin grinding wheels based on an AI model, characterized in that, include: Step S1: Simultaneously acquire visible light grayscale characteristics and near-infrared light intensity characteristics of the resin grinding wheel under test while it is rotating; Step S2: Align the visible light grayscale feature information with the near-infrared light intensity feature information in spatial coordinates to construct a multispectral mapping field that reflects the physical properties of the surface and subsurface of the resin grinding wheel under test. Step S3: Based on the AI ​​model, identify the particle size parameters of the resin grinding wheel to be tested, and according to the sampling window size determined by the particle size parameters, perform nonlinear correlation calculation by combining the visible light grayscale feature information, the near-infrared light intensity feature information and the grayscale mean value within the sampling window to obtain the local effective abrasive concentration projection value. Step S4: Extract the local effective abrasive concentration projection value of each coordinate point in the full circumference, calculate the dispersion parameter of the local effective abrasive concentration projection value in the preset area, and output the uniformity detection conclusion of the resin grinding wheel to be tested based on the comparison result of the dispersion parameter and the preset threshold.

2. The method for detecting the uniformity of resin grinding wheels based on an AI model according to claim 1, characterized in that, The calculation of the local effective abrasive concentration projection value in step S3 is specifically as follows: multiply the near-infrared light intensity feature value and the visible light grayscale feature value at a specific coordinate point to obtain a product term, and divide the product term by the square of the mean visible light grayscale value within the sampling window to obtain the local effective abrasive concentration projection value.

3. The method for detecting the uniformity of resin grinding wheels based on an AI model according to claim 1, characterized in that, The step S4 of calculating the dispersion parameter of the local effective abrasive concentration projection value within the preset area specifically involves calculating the variance of the local effective abrasive concentration projection value corresponding to all coordinate points in the full circumference.

4. The method for detecting the uniformity of resin grinding wheels based on an AI model according to claim 1, characterized in that, In step S4, the uniformity detection conclusion of the resin grinding wheel under test is output based on the comparison result of the dispersion parameter and the preset threshold. Specifically, when the variance of the local effective abrasive concentration projection value in the whole circumference is greater than the preset threshold, it is determined that the resin grinding wheel under test has an internal abrasive uneven distribution defect.

5. A uniformity detection system for resin grinding wheels based on an AI model, characterized in that, The method for detecting the uniformity of resin grinding wheels based on an AI model, as described in any one of claims 1 to 4, comprises: The feature information acquisition module simultaneously acquires the visible light grayscale feature information and near-infrared light intensity feature information of the resin grinding wheel under test in a rotating state; The spatial coordinate alignment module aligns the visible light grayscale feature information with the near-infrared light intensity feature information in spatial coordinates to construct a multispectral mapping field that reflects the physical properties of the surface and subsurface of the resin grinding wheel under test. The projection value calculation module identifies the particle size parameters of the resin grinding wheel under test based on the AI ​​model, and performs nonlinear correlation calculation based on the sampling window size determined by the particle size parameters, combined with the visible light grayscale feature information, the near-infrared light intensity feature information, and the grayscale mean value within the sampling window, to obtain the local effective abrasive concentration projection value. The uniformity detection module extracts the local effective abrasive concentration projection values ​​at each coordinate point along the entire circumference, calculates the dispersion parameter of the local effective abrasive concentration projection values ​​within a preset area, and outputs the uniformity detection conclusion of the resin grinding wheel under test based on the comparison result of the dispersion parameter and a preset threshold.

Images