A computer memory bar aging test system

Abstract

The present invention discloses a computer memory bar aging test system, which relates to the field of computer hardware testing technology. The system divides the surface of a PCB into grids and silk-screens coordinate reference points. A thermal imager is used to collect the temperature of each grid in real time, constructing a sequence of average temperature and smoothed grayscale values. The standard deviation of the amplitude spectrum of the smoothed grayscale values is extracted through FFT transformation, and the ASI (Asymmetric Scaling Index) is calculated by combining kurtosis and skewness. A heat map is generated based on the aging degree of different grids. A three-stage ASI change rate sequence is analyzed based on an LSTM model to determine the life cycle stage and aging type of the computer memory bar. The system improves the accuracy of memory bar aging assessment through multi-dimensional data fusion, reducing the maintenance cost of the memory bar and improving system reliability.

Description

Technical Field

[0001] The present invention relates to the technical field of computer hardware testing, in particular to a computer memory bar aging testing system. Background Art

[0002] With the continuous advancement of computer technology, computer memory modules are becoming increasingly important in computer systems. The aging of memory modules directly affects computer performance and stability. However, there is currently a lack of efficient and accurate methods to directly detect the aging of memory module components. Existing detection methods mostly rely on complex disassembly and specialized equipment, which is cumbersome and may damage the memory module. Therefore, the development of a test system that can accurately assess the aging of memory modules through a non-invasive and indirect method is of great practical significance. Summary of the Invention

[0003] In response to the shortcomings of the existing technology, the present invention provides a computer memory bar aging test system, which solves the problems of the existing technology that the operation is cumbersome and the memory bar is easily damaged, and the aging degree of the memory bar cannot be non-invasively, accurately and in real time evaluated.

[0004] To achieve the above objectives, the present invention is implemented through the following technical solutions: A computer memory bar aging test system, comprising:

[0005] The grid positioning module divides the surface of the outermost PCB board of the computer memory module into a square grid of uniform area, assigns unique coordinates to each grid, prints grid coordinate reference points on the edge of the PCB board, and transmits the divided coordinate grid to the grid temperature acquisition module;

[0006] The grid temperature acquisition module uses a thermal imager to collect thermal images within T minutes at a frequency of f1, calculates the average temperature of each grid in each thermal image, combines all images within T minutes to obtain the average temperature sequence corresponding to each grid, and transmits the average temperature sequence to the grid aging calculation module;

[0007] The grid aging calculation module converts the average temperature sequence of each grid into a grayscale value sequence and performs smoothing to obtain the smoothed grayscale value sequence Gt. The aging sensitivity index ASI is constructed based on the kurtosis, skewness, and amplitude spectrum standard deviation of Gt, and the ASI of each grid is transmitted to the thermal map generation module.

[0008] The heat map generation module maps the grids to different colors and levels based on their ASI values, uses the coordinates corresponding to each color grid to generate a heat map, and transmits the heat map to the memory bar aging judgment module;

[0009] The memory module for aging judgment generates heat maps with the same grid division for three stages of the memory life cycle respectively, calculates the change rate of each grid compared with the reference ASI in the three stages, obtains different grid ASI change rate sequences A1i, A2i, A3i, substitutes multiple groups of A1i, A2i, A3i into the long short-term memory network for training, inputs the ASI change rate sequence of the untested memory into the trained model to determine the life cycle stage of the memory, calculates the proportions Green and Red of the green grids and red grids according to the heat map, and combines the distribution of the red grids to judge the aging type. Calculate the values of Green and Red by counting the number of all green grids and red grids in the heat map and then dividing by the total number of grids respectively. The three stages of the memory life cycle are the initial stage, the middle stage, and the end stage.

[0010] As a further solution of the present invention, the average temperature sequence is converted into a grayscale value sequence G according to the formula G=(T1-a)×(255-0)÷(b-a), where a and b are respectively the minimum value and the maximum value of the average temperature sequence, and T1∈[a,b].

[0011] As a further solution of the present invention, a fast Fourier transform is performed on Gt to convert the time-domain signal into a frequency-domain signal F(f), obtain the amplitude spectrum |F(fhight)| of the frequencies greater than or equal to flimit, and calculate its standard deviation, where flimit is the aging-sensitive frequency threshold.

[0012] As a further solution of the present invention, the aging degree of the grid is calculated according to the formula ASI=piu*((kurtosis(Gt)+skewness(Gt) / 2)), where piu is the standard deviation of the amplitude spectrum.

[0013] As a further solution of the present invention, based on the ASI distribution of multiple groups of samples with different aging degrees, the Otsu algorithm is used to automatically determine the thresholds Grade1, Grade2, Grade3, and the aging grades of the grids are divided as follows: ASI<=Grade1, healthy, classified as grade 1, corresponding to green; Grade1<ASI<=Grade2, slightly aging, classified as grade 2, corresponding to blue; Grade2<ASI<=Grade3, moderately aging, classified as grade 3, corresponding to white; ASI>Grade3, severely aging, classified as grade 4, corresponding to red.

[0014] As a further solution of the present invention, the specific steps for generating the heat map are as follows:

[0015] Map the ASI value of each grid to the corresponding color grade;

[0016] Bind the coordinates of each grid to its corresponding color grade;

[0017] Calibrate the grid coordinates to the actual physical positions based on the silkscreen reference points on the edge of the PCB board;

[0018] Create a canvas that matches the size of the PCB board, traverse each grid, fill it with the corresponding color according to its aging level, and mark the grid coordinate reference points and scale lines.

[0019] As a further embodiment of the present invention, according to the formula Aqi=(ASIq-ASI 基准 ) / ASI 基准 Calculate the ASI change rate of the grid at different stages, where ASIq is the grid ASI of the current memory bank, and ASI 基准 Obtained for testing new memory sticks that have not aged.

[0020] As a further solution of the present invention, the specific steps of determining the aging type of the memory stick are:

[0021] If Red<=th1 and Green>=gh2, the memory module is judged to be in a state of continuous and stable aging;

[0022] If Red∈(th1,th2], Green∈[gh1,gh2] and the red grids are scattered, the memory bar is judged to be in intermittent aging;

[0023] If Red∈(th1,th2], Green∈[0,gh1] and the red grids are clustered, the memory bar is judged to be in accelerated aging;

[0024] If Red>th2, the memory module is judged to be seriously aged and a mandatory warning is required. th1, th2, gh1, and gh2 are the percentage thresholds.

[0025] As a further solution of the present invention, the specific method for judging whether the red grid is gathered or dispersed is as follows:

[0026] Taking the red grid as the center, for each red grid, count the 9 grids in the surrounding 3×3 neighborhood, including the red grid itself, count the number of red grids in these 9 grids, and then divide by 9 to get the red percentage of the neighborhood;

[0027] If there is at least one neighborhood with a red ratio greater than or equal to 30%, the red grid is considered concentrated;

[0028] If the red proportion of all neighborhoods is less than 30%, the red grid is judged to be scattered.

[0029] The present invention provides a computer memory module aging test system. Compared with the prior art, it has the following advantages:

[0030] (1) The present invention uses non-invasive thermal imaging technology to dynamically monitor the aging process of memory modules in real time without disassembling them. The aging area is located through fine grid division, which can accurately capture abnormal temperature fluctuations and significantly improve the efficiency and accuracy of aging detection.

[0031] (2) The present invention constructs an aging sensitivity index model by extracting the temporal and spatial characteristics of the grid temperature sequence, intuitively presents the aging distribution with the help of heat map visualization technology, intelligently judges the life cycle stage by combining the LSTM model, and collects the proportion of grids of different colors through the heat map to judge the aging stage of the memory stick in different life cycles, providing quantitative data support for the full life cycle management of the memory stick. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is a block diagram of the system principle of the present invention. DETAILED DESCRIPTION

[0033] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0034] like Figure 1 The present invention provides a computer memory bar aging test system, comprising:

[0035] The grid positioning module divides the surface of the outermost PCB board of a computer memory module into a uniform square grid. Each grid has the same area, facilitating subsequent unified data collection and analysis of each area. The grid size can be adjusted based on the actual size of the PCB board and the test accuracy requirements. For example, a DDR4 memory module PCB board with a length of 133.35mm and a width of 31.75mm can be divided into a square grid with a side length of 5mm. The length direction can be divided into approximately 133.35÷5≈27 grids, and the width direction can be divided into approximately 31.75÷5≈6 grids, for a total of approximately 27×6=162 grids.

[0036] Assign unique coordinates (x, y) to each grid. The coordinates are set using the Cartesian coordinate system, with the upper left corner of the PCB board as the origin (0, 0). The positive direction of the x-axis is horizontally to the right, and the positive direction of the y-axis is vertically downward. For example, in the above-described grid division, the coordinates of the first grid in the first row are (0, 0), the coordinates of the second grid in the first row are (1, 0), and the coordinates of the first grid in the second row are (0, 1). And so on. Each grid can be accurately identified by a unique coordinate, which facilitates subsequent data processing and positioning.

[0037] Grid coordinate reference points are silk-screened on the edge of the PCB board to facilitate operators to quickly locate each grid in actual operation. For example, we can silk-screen (0,0) in the upper left corner of the PCB board, (26,0) in the upper right corner, (0,5) in the lower left corner, and (26,5) in the lower right corner. Operators can use these reference points to quickly find the corresponding grid position when viewing thermal imaging images or performing data analysis.

[0038] The grid temperature acquisition module uses a thermal imager to collect thermal images at a frequency of f1 for T minutes. The acquisition frequency f1 and acquisition duration T need to be set according to specific test requirements. For example, for a general memory bar burn-in test, we can set the acquisition frequency f1 to 5Hz, which means that 5 thermal images are collected per second. The acquisition duration T can be set to 10 minutes, or 600 seconds. During these 10 minutes, the thermal imager will collect a total of 600 × 5 = 3000 thermal images.

[0039] Collect the average temperature of each grid in each thermal image. The images captured by the thermal imager contain temperature distribution information on the surface of the PCB board. We can use image processing and data analysis techniques to extract the temperature value within each grid and calculate its average value. For example, a grid in a thermal image may have multiple pixels inside it, and each pixel corresponds to a temperature value. We can add the temperature values of these pixels and divide it by the number of pixels to get the average temperature of the grid. By collecting the average temperature of each grid in 3000 thermal images, we can obtain the average temperature sequence A1, A2, ..., A3000 of each grid within 10 minutes. Assume that the average temperature of a grid in the first second is 32°C, 32.5°C in the second second, and so on, forming a temperature sequence [32, 32.5, ...].

[0040] Grid aging calculation module, which converts the average temperature sequence of each grid into a grayscale value sequence. The conversion can be performed using a linear mapping method, that is, converting the values in the average temperature sequence into the range of 0-255. This step is to facilitate subsequent signal processing and analysis. For example, assuming that the average temperature sequence of a grid is [20℃,...,60℃], 20℃ needs to be mapped to a grayscale value of 0 and 60℃ needs to be mapped to a grayscale value of 255. Then, for a temperature value T1 of the grid, its corresponding grayscale value G can be calculated by the following formula: G=(T1-20)×(255-0)÷(60-20). If T=30℃, then G=(30-20)×(255-0)÷(60-20)=63.75, which is rounded to 64.

[0041] The average temperature sequence is divided into sliding windows of t seconds, and the grayscale average of each window is calculated to obtain the smoothed grayscale sequence G1,...,Gm. The size of the sliding window t needs to be adjusted according to the actual situation. Generally, 5-10 seconds is selected. For example, we choose a sliding window of t=5 seconds. For the 3000 average temperature data points collected earlier, they correspond to 3000 grayscale values. With 5 seconds as a window, each window contains 5×5=25 grayscale values. We calculate the average of the 25 grayscale values in each window and use it as the grayscale average corresponding to the window. The grayscale average of the first window is recorded as G1, the grayscale average of the second window is recorded as G2, and so on, until the grayscale average of all windows is obtained to form a smoothed grayscale sequence.

[0042] Perform FFT (Fast Fourier Transform) on the smoothed grayscale sequence Gt to convert the time domain signal into a frequency domain signal F(f). The FFT transform can decompose the signal into components of different frequencies, facilitating analysis of the signal's frequency characteristics. Define the aging-sensitive frequency range fhight>=flimit. Here, flimit is a threshold determined through extensive experiments and data analysis. For example, through testing and analysis of multiple groups of aged memory sticks, we determined the aging-sensitive frequency range to be 15 Hz and above, i.e., flimit=15. Extract the amplitude spectrum |F(fhight)| of this segment and calculate its standard deviation piu. The standard deviation reflects the degree of data dispersion. A larger piu value indicates greater amplitude fluctuations within this frequency range, potentially indicating more severe aging of the grid.

[0043] Calculate the kurtosis and skewness of Gt. The kurtosis describes the peak degree of the data distribution, and the skewness describes the asymmetry of the data distribution. If the kurtosis of the grayscale sequence of a grid is large, it means that the data distribution is relatively peaked, and there may be some abnormal temperature fluctuations. If the skewness is not 0, it means that the data distribution is asymmetric, which may indicate the unevenness of the aging process.

[0044] Using only the standard deviation of the amplitude spectrum may not be able to comprehensively capture the aging characteristics. For example, the standard deviations of two grids are the same, but one has a high kurtosis indicating more extreme fluctuations, while the other has a low kurtosis indicating uniform fluctuations. The former is more likely to be in an aging state; and skewness can reflect the long-term trend of temperature change, such as a gradual increase, which cannot be reflected by the standard deviation;

[0045] Therefore, starting from the time domain and frequency domain of Gt, combine the kurtosis, skewness, and standard deviation of the amplitude spectrum of Gt to construct an aging sensitivity index ASI. The specific formula is: ASI = piu * ((kurtosis(Gt) + skewness(Gt) / 2)). Through this formula, an index that can reflect the aging degree of the grid is obtained. For example, if piu = 0.8, kurtosis(Gt) = 4, and skewness(Gt) = 1 for a certain grid, then the ASI of this grid = 0.8×((4 + 1 / 2)) = 3.6;

[0046] Based on the ASI distribution of multiple groups of samples with different aging degrees, use the Otsu algorithm to automatically determine the threshold. The Otsu algorithm is a commonly used image segmentation algorithm. It can automatically find an optimal threshold according to the distribution of data and divide the data into two categories. For example, by analyzing 200 groups of memory module samples with different aging degrees, use the Otsu algorithm to determine Grade1 = 1.5, Grade2 = 2.5, Grade3 = 3.5. According to these thresholds, we can divide the aging level of the grid into: ASI <= Grade1, healthy; Grade1 < ASI <= Grade2, slightly aging; Grade2 < ASI <= Grade3, moderately aging; ASI > Grade3, severely aging.

[0047] The heat map generation module divides the grid into 4 levels according to the ASI value. ASI <= Grade1 is classified as level 1, corresponding to green; Grade1 < ASI <= Grade2 is classified as level 2, corresponding to blue; Grade2 < ASI <= Grade3 is classified as level 3, corresponding to white; ASI > Grade3 is classified as level 4, corresponding to red; The selected color correspondence method is based on people's intuitive perception of colors. Green usually represents healthy and normal, and red represents dangerous and severe. For example, for the grid with ASI = 3.6 calculated above, since 3.6 > 3.5, this grid belongs to level 4 and corresponds to red;

[0048] Map the ASI value of each grid to the corresponding color level; bind the coordinates (x, y) of each grid to its corresponding color level to form structured data, such as a dictionary or two-dimensional array; calibrate the virtual grid coordinates to the actual physical location based on the silkscreen reference points on the edge of the PCB board; draw grid lines to ensure alignment with the circuit layout of the PCB board; create a canvas that matches the size of the PCB board, traverse each grid, fill it with the corresponding color according to its aging level, and mark the grid coordinate reference points and scale lines.

[0049] The memory stick aging judgment module divides the memory stick life cycle into three stages: initial stage, mid-stage, and final stage. In each stage, we generate an aging heat map with the same grid division according to the previous method and calculate the change rate of each grid compared to the baseline ASI in the three stages. For example, taking one of the grids as an example, the baseline ASI is obtained by testing a grid of the same batch of new memory sticks. 基准 =0.6. In the initial stage, the ASI of a certain grid is 1; in the middle stage, the ASI of the grid becomes 1.8; in the final stage, the ASI of the grid becomes 3. Then the ASI change rate of the grid relative to the baseline in the initial stage is (1-0.6) / 0.6≈0.67, the ASI change rate relative to the baseline in the middle stage is (1.8-0.6) / 0.6=2, and the ASI change rate relative to the baseline in the final stage is (3-0.6) / 0.6=4. By performing this calculation for all grids, we obtain the ASI change rate sequence A1i, A2i, A3i in each stage, i∈[1,M], where M is the total number of heat maps obtained in each stage;

[0050] Multiple sets of A1i, A2i, and A3i sequences at different stages are fed into an LSTM (long short-term memory) network for training. LSTM is a special recurrent neural network that can process sequential data and has a memory function. It is suitable for analyzing time series information during the aging process of memory sticks. For example, we can use deep learning frameworks such as TensorFlow or PyTorch to build and train LSTM models. During training, we use memory stick data at known life cycle stages as the training set and adjust the model parameters so that the model can accurately determine the life cycle stage of the memory stick based on the ASI change rate sequence. The trained model can then predict new memory stick data and determine its current life cycle stage.

[0051] After determining the current lifecycle stage of the memory stick, calculate the percentage of green and red grids based on the heat map. Count the number of green grids and the number of red grids, then divide them by the total number of grids to obtain the Green and Red values. Based on the Green and Red values and the distribution of red grids, determine the aging type.

[0052] If Red <= 5% and Green >= 60%, the memory module is in stable aging. This means that most areas of the memory module are healthy, with only a few areas showing slight aging. The aging process is slow and stable.

[0053] If Red∈[5%,15%], Green∈[40%,60%], and the red grids are scattered, it is judged to be intermittent aging, indicating that a certain proportion of the memory module has aged, but the aging areas are relatively scattered, which may be due to local aging caused by some accidental factors;

[0054] If Red∈[5%,15%], Green∈[0,40%], and the red grids are clustered, it is judged to be in accelerated aging. In this case, the aging areas not only have a certain proportion but are also clustered together, indicating that the aging process is accelerating, which may have a significant impact on the performance of the memory stick.

[0055] If Red > 15%, a mandatory warning is issued, indicating that the memory module is severely aged and may fail at any time, requiring timely replacement or repair.

[0056] The method for determining whether red grids are clustered or dispersed is as follows: with each red grid as the center, calculate the red proportion in the 3×3 neighborhood around it. For example, for a certain red grid, its surrounding 3×3 neighborhood contains 9 grids, including the red grid itself. Count the number of red grids in these 9 grids, and then divide by 9 to get the red proportion of the neighborhood. If there is at least one neighborhood with a red proportion greater than or equal to 30%, the red grid is judged to be concentrated. If the red proportion of all neighborhoods is less than 30%, the red grid is judged to be dispersed. This method can accurately determine the distribution of red grids, and thus more accurately determine the aging type.

[0057] Some of the data in the above formulas are dimensionless and numerically calculated. Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0058] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A computer memory bar aging test system, characterized in that: include: The grid positioning module divides the surface of the outermost PCB board of the computer memory module into a square grid of uniform area, assigns unique coordinates to each grid, prints grid coordinate reference points on the edge of the PCB board, and transmits the divided coordinate grid to the grid temperature acquisition module; The grid temperature acquisition module uses a thermal imager to collect thermal images within T minutes at a frequency of f1, calculates the average temperature of each grid in each thermal image, combines all images within T minutes to obtain the average temperature sequence corresponding to each grid, and transmits the average temperature sequence to the grid aging calculation module; The grid aging calculation module converts the average temperature sequence of each grid into a grayscale value sequence and performs smoothing to obtain the smoothed grayscale value sequence Gt. The aging sensitivity index ASI is constructed based on the kurtosis, skewness, and amplitude spectrum standard deviation of Gt, and the ASI of each grid is transmitted to the thermal map generation module. The heat map generation module maps the grids to different colors and levels based on their ASI values, uses the coordinates corresponding to each color grid to generate a heat map, and transmits the heat map to the memory bar aging judgment module; The memory stick aging judgment module generates heat maps with the same grid division for the three stages of the memory stick life cycle, calculates the change rate of each grid in the three stages compared with the baseline ASI, and obtains the ASI change rate sequences A1i, A2i, and A3i of different grids. Multiple groups of A1i, A2i, and A3i are substituted into the long short-term memory network for training. The ASI change rate sequences of untested memory sticks are input into the trained model to determine the life cycle stage of the memory stick. The proportions of green and red grids (Green and Red) are calculated based on the heat map, and the aging type is determined based on the distribution of red grids. The Green and Red values are calculated by counting all green grids and red grids in the heat map and then dividing them by the total number of grids. The three stages of the memory stick life cycle are the initial stage, the middle stage, and the final stage.

2. A computer memory bar aging test system according to claim 1, characterized in that: The average temperature sequence is converted into a grayscale value sequence G according to the formula G=(T1-a)×(255-0)÷(ba), where a and b are the minimum and maximum values of the average temperature sequence, respectively, and T1∈[a,b].

3. A computer memory bar aging test system according to claim 1, characterized in that: Perform a fast Fourier transform on Gt to convert the time domain signal into a frequency domain signal F(f). Obtain the amplitude spectrum |F(fhight)| for frequencies greater than or equal to flimit, and calculate its standard deviation, where flimit is the aging-sensitive frequency threshold.

4. A computer memory bar aging test system according to claim 3, characterized in that: The aging degree of the grid is calculated according to the formula ASI=piu*((kurtosis(Gt)+skewness(Gt) / 2)), where piu is the standard deviation of the amplitude spectrum.

5. The computer memory bar aging test system according to claim 1, characterized in that: Based on the ASI distributions of multiple groups of samples with different aging degrees, the Otsu algorithm is used to automatically determine the thresholds Grade1, Grade2, and Grade3, and the aging grades of the grids are classified as follows: ASI <= Grade1, healthy, classified as Grade 1, corresponding to green; Grade1 < ASI <= Grade2, mildly aged, classified as Grade 2, corresponding to blue; Grade2 < ASI <= Grade3, moderately aged, classified as Grade 3, corresponding to white; ASI > Grade3, severely aged, classified as Grade 4, corresponding to red.

6. A computer memory bar aging test system according to claim 1, characterized in that: The specific steps to generate the heat map are as follows: Map the ASI value of each grid to the corresponding color grade; Bind the coordinates of each grid to its corresponding color grade; Calibrate the grid coordinates with the actual physical position according to the silk screen reference points on the edge of the PCB board; Create a canvas that matches the size of the PCB board, traverse each grid, fill in the corresponding color according to its aging grade, and mark the grid coordinate reference points and scale lines.

7. A computer memory bar aging test system according to claim 1, characterized in that: According to the formula Aqi=(ASIq-ASI 基准 ) / ASI 基准 Calculate the ASI change rate of the grid at different stages, where ASIq is the grid ASI of the current memory bank, and ASI 基准 Obtained for testing new memory sticks that have not aged.

8. The computer memory bar aging test system according to claim 1, characterized in that: The specific steps to determine the aging type of the memory module are as follows: If Red <= th1 and Green >= gh2, it is determined that the memory module is in continuous and stable aging; If Red ∈ (th1, th2], Green ∈ [gh1, gh2) and the red grids are scattered, it is determined that the memory module is in intermittent aging; If Red ∈ (th1, th2], Green ∈ [0, gh1) and the red grids are aggregated, it is determined that the memory module is in accelerated aging; If Red > th2, it is determined that the memory module is severely aged and a forced warning is required, where th1, th2, gh1, and gh2 are proportion thresholds.

9. A computer memory bar aging test system according to claim 8, characterized in that: The specific method to determine the aggregation and dispersion of the red grids is as follows: Taking the red grid as the center, for each red grid, count the 9 grids in the surrounding 3×3 neighborhood, including the red grid itself, count the number of red grids in these 9 grids, and then divide by 9 to obtain the red proportion in this neighborhood; If there is at least one neighborhood with a red proportion greater than or equal to 30%, it is determined that the red grids are concentrated; If the red proportion of all neighborhoods is less than 30%, it is determined that the red grids are scattered.

Images