System and method for repairing voids in a substrate

Abstract

This invention provides a system and method for repairing defects on a substrate. The system includes a transfer device, an image recognition module, and a processing module. The transfer device is configured to transfer at least one first chip on a first substrate to at least one defect on a second substrate. The image recognition module views and obtains a first position data of the at least one first chip on the first substrate and a second position data of the at least one second chip on the second substrate. The processing module inputs the first position data and the second position data into a predictive machine model to obtain an optimal relative displacement. The transfer device generates a motion with the optimal relative displacement between the second substrate and the first substrate to transfer the at least one first chip.

Description

Technical Field

[0001] This invention relates to a system and method for repairing defects on a substrate, and more particularly to a system and method for repairing defects in chips on a target substrate using a machine learning model. Background Technology

[0002] Mass transfer is a key technology for the mass production of micro LED displays. Millions or even tens of millions of micrometer-sized RGB Micro LED chips are transferred in batches from the original substrate to a display substrate containing the driving circuitry using mass transfer technology. After being irradiated and energized by a laser beam, the release layer material absorbs the energy, causing the release layer between the chip and the temporary substrate or carrier substrate to lose its adhesion. Once the chip is peeled off from the release layer, it moves towards the target substrate, completing the transfer.

[0003] However, after mass transfer, numerous chip gaps may appear on the target substrate (or the display's driving substrate). Therefore, a repair device is needed to implant chips into the locations of these gaps. To repair the gaps, the repair device needs to move the temporary substrate or the target substrate so that the chips on the temporary substrate are aligned with the gaps on the target substrate, and the chip transfer is completed by irradiating with a laser beam.

[0004] Because the gaps on the target substrate are irregularly distributed, and gaps may also form on the chips on the temporary substrate after transfer, or some chips may be deemed abnormal or defective and cannot be implanted into the target substrate, the relative displacement between the temporary substrate and the target substrate should be minimized each time, and the number of chips on the temporary substrate aligned with the gaps on the target substrate should be maximized. This can increase the transfer efficiency of the repair equipment. Summary of the Invention

[0005] In one embodiment of the present invention, a system for repairing a gap on a substrate is provided, comprising: a transfer device configured to transfer at least one first chip on a first substrate to at least one gap on a second substrate; an image recognition module for viewing and obtaining a first position data of the at least one first chip on the first substrate and a second position data of the at least one second chip on the second substrate; and a computational processing module for inputting the first position data and the second position data into a predictive machine model to obtain an optimal relative displacement; wherein the transfer device generates a movement between the second substrate and the first substrate with the optimal relative displacement to transfer the at least one first chip.

[0006] In another embodiment of the invention, the optimal relative displacement is obtained by adjusting the parameters in the predictive machine model to reduce the loss value.

[0007] In another embodiment of the present invention, after the transfer device completes the transfer of at least one first chip, it calculates the number of vacancies still existing on the second substrate.

[0008] In another embodiment of the present invention, if the number of remaining vacancies is not zero, the first position data and the second position data are updated, and another optimal relative displacement required for the next round of chip transfer is calculated.

[0009] In another embodiment of the present invention, the architecture for forming the predictive machine model includes: a repair platform simulator that generates virtual first position data and virtual second position data; a platform displacement prediction model that takes the virtual first position data and the virtual second position data as input to calculate multiple platform displacement scores, selects the optimal virtual relative displacement corresponding to the highest score from the platform displacement scores, and provides the optimal virtual relative displacement to the repair platform simulator to generate simulated motion and simulated transfer chips, and outputs a transfer quantity; and a loss function model that updates the parameters in the platform displacement prediction model according to the transfer quantity and the platform displacement scores.

[0010] In one embodiment of the present invention, a method for repairing a gap on a substrate is provided, comprising: examining at least one first chip on a first substrate to obtain a first position data and examining at least one second chip on a second substrate to obtain a second position data; inputting the first position data and the second position data into a predictive machine model to obtain an optimal relative displacement; generating a movement between the second substrate and the first substrate having the optimal relative displacement; and transferring the at least one first chip to the at least one gap.

[0011] In another embodiment of the present invention, the step of obtaining the optimal relative displacement is to reduce the loss value by adjusting the parameters in the predictive machine model, thereby obtaining the optimal relative displacement.

[0012] In another embodiment of the invention, the method further includes calculating the number of vacancies still existing on the second substrate after the transfer of the at least one first chip is completed.

[0013] In another embodiment of the invention, if the number of remaining vacancies is not zero, the first position data and the second position data are updated, and another optimal relative displacement required for the next round of chip transfer is calculated.

[0014] In another embodiment of the present invention, the predictive machine model is formed by the following steps: generating a virtual first position data and a virtual second position data; inputting the virtual first position data and the virtual second position data to calculate a plurality of platform displacement fractions; selecting an optimal virtual relative displacement corresponding to the highest score from the platform displacement fractions; simulating the motion between a virtual first substrate and a virtual second substrate and simulating the transfer chip with the optimal virtual relative displacement, and outputting a transfer quantity; and updating the parameters in the predictive machine model according to the transfer quantity and the platform displacement fractions. Attached Figure Description

[0015] To fully understand the nature, advantages, and preferred embodiments of the present invention, the following detailed description can be more clearly understood by referring to the accompanying drawings.

[0016] Figure 1 A block architecture diagram of a system for repairing gaps on a substrate is shown according to an embodiment of the present invention.

[0017] Figure 2A and 2B This is a schematic diagram depicting the chip distribution and state on a carrier substrate and a target substrate according to an embodiment of the present invention.

[0018] Figure 2C and Figure 2D To depict Figure 2A and Figure 2B A schematic diagram showing the location data of chip distribution and status on the carrier substrate and the target substrate.

[0019] Figure 3 A schematic diagram illustrating the relative displacement of the carrier substrate with respect to the target substrate.

[0020] Figure 4 A schematic diagram illustrating the architecture of a machine learning model that can plan repair paths.

[0021] Figure 5 A flowchart illustrating the training of a predictive machine model capable of planning repair paths.

[0022] Figure 6 This is a diagram illustrating the architecture of the platform displacement prediction model.

[0023] Figure 7 A flowchart illustrating the patching of a predictive machine model.

[0024] Figure 8A and Figure 8B This is a schematic diagram depicting the chip distribution on a carrier substrate and a target substrate according to an embodiment of the present invention.

[0025] Figures 8C to 8FA schematic diagram depicting four possible relative displacements between the carrier substrate and the target substrate.

[0026] Explanation of key component symbols:

[0027] 10: System

[0028] 11: Calculation and Processing Module

[0029] 12: Control Module

[0030] 13: Transfer device

[0031] 14: Image Recognition Module

[0032] 15: Laser source

[0033] 40: Architecture

[0034] 41: Patch Platform Simulator

[0035] 42: Platform Displacement Prediction Model

[0036] 43: Loss Function Model

[0037] 81: Carrier substrate

[0038] 82: Target substrate

[0039] 131: Servo Motor Module

[0040] 132: Carrier Platform

[0041] 133: Target Platform

[0042] 421: Machine Learning Model

[0043] 422: Retrieve the relative displacement corresponding to the highest score

[0044] 501-507: Steps

[0045] 701-705: Steps

[0046] 811: Chip

[0047] 822: Chip

[0048] 4211: Parameter P

[0049] 81a: Carrier substrate

[0050] 81d: Defective chip

[0051] 81v: Vacant

[0052] 82b: Target substrate

[0053] 82v: Vacant

[0054] A: Relative displacement

[0055] C: Reference point

[0056] P: Parameters

[0057] Q: Platform displacement fraction

[0058] R: Quantity

[0059] T: Reference point

[0060] S1, S2: Array Detailed Implementation

[0061] The following description illustrates preferred embodiments of the present invention. The invention will be described below with reference to the embodiments and accompanying drawings. Therefore, the invention is not intended to be limited to the embodiments shown, but rather to conform to the principles disclosed herein. Furthermore, those skilled in the art will make various modifications or variations based on the present invention, and these will be incorporated into the spirit and scope of this document and the appended claims.

[0062] Figure 1 This diagram illustrates the block architecture of a system for repairing gaps on a substrate. The system 10 in this embodiment includes a processing module 11, a control module 12, a transfer device 13, an image recognition module 14, and a laser source 15. In this embodiment, the transfer device 13 includes a servo motor module 131, a carrier platform 132, and a target platform 133. The servo motor module 131 generates relative movement between the carrier platform 132 and the target platform 133 on a horizontal plane, aligning specific positions of the carrier substrate 81 (or first substrate) on the carrier platform 132 and the target substrate 82 (or second substrate) on the target platform 133. Thus, after one or more chips 811 (or first chips) on the carrier substrate 81 are irradiated by a beam emitted from the laser source 15 onto a release layer (not shown), they are implanted into the gaps between chips 822 (or second chips) on the target substrate 82.

[0063] The image recognition module 14 includes an image acquisition unit and a discrimination unit, which can respectively inspect the distribution and status (whether normal or defective) of chips (811, 822) on the carrier substrate 81 and the target substrate 82, and obtain the first position data of chip 811 on the carrier substrate 81 and the second position data of chip 822 on the target substrate 82. The first position data and the second position data include the positions of the chips and vacancy, as well as the status of the chips. The servo motor module 131 can move the carrier platform 132 and the target platform 133 respectively according to the control signal generated by the control module 12, thereby obtaining a better or optimal relative displacement. This relative displacement is calculated and output by the arithmetic processing module 11, which is based on the first position data and the second position data input into a predictive machine model, and by adjusting the parameters in the predictive machine model to reduce the loss value, an optimal relative displacement vector is obtained. The training and learning of the predictive machine model, and the execution of the transfer chip calculation process by the system 10, will be discussed in subsequent paragraphs.

[0064] Figure 2A and Figure 2B This is a schematic diagram depicting the chip distribution and state on a carrier substrate and a target substrate according to an embodiment of the present invention. After inspection and judgment by the image recognition module 14, chip 811, defective chip 81d, and void 81v are found on the carrier substrate 81, and chip 822 and void 82v are found on the target substrate 82. Based on these image judgment results, they can be converted into first position data and second position data, represented in an array, which the arithmetic processing module 11 needs. See also... Figure 2C The first position data can be represented as a 5×5 matrix, which is then further converted into a 25×1 array S1. Similarly, the second position data can be represented as a 4×4 matrix, which is then further converted into a 16×1 array S2. These arrays S1 and S2 are only examples; the actual carrier substrate and target substrate have a much larger number of chips.

[0065] Figure 3 This is a schematic diagram illustrating the relative displacement of the carrier substrate relative to the target substrate. For example, with the upper left corner of the carrier substrate 81 as reference point C and the upper left corner of the target substrate 82 as reference points T, the relative displacement A can be expressed as the displacement of reference point C relative to reference point T, for example, represented by the displacement vector {Ax, Ay} to show the direction and amount of movement. The direction and magnitude of the relative displacement A in the figure are merely illustrative and do not limit the scope of this application.

[0066] The architecture for training and learning predictive machine models is as follows: Figure 4As shown, this is so that it can be applied to the aforementioned system 10 to plan the optimal repair path. Architecture 40 includes a repair platform simulator 41, a platform displacement prediction model 42, and a loss function model 43. The repair platform simulator 41 includes a virtual carrier platform, a virtual target platform, and a virtual laser source. In the calculation, the repair platform simulator 41 generates virtual arrays (S1, S2) (see...). Figure 2C and Figure 2D The platform displacement prediction model 42 is used to obtain an optimal relative displacement A. The virtual carrier platform and virtual target platform in the repair platform simulator 41 will simulate motion based on the optimal relative displacement A. The virtual laser source will transfer chips from the virtual carrier platform to vacancies on the virtual target platform. After the virtual laser source completes the chip transfer sequentially, the repair platform simulator 41 outputs the number of transferred chips R to the loss function model 43. The loss function model 43 updates the parameter P in the platform displacement prediction model 42 based on the number R and the platform displacement score Q. The platform displacement prediction model 42 then uses the updated parameter P to calculate the relative displacement A in the next round, and selects the relative displacement A (optimal) corresponding to the highest platform displacement score Q before outputting it to the repair platform simulator 41.

[0067] Figure 5 A flowchart is provided to illustrate the training of a predictive machine model capable of planning repair paths. As shown in step 501, at the beginning of the first round of training and learning (i.e., N=1), the repair platform simulator 41 generates virtual initial arrays S1(1) and S2(1) representing first position data and second position data, and sets the number of transferred chips R(1)=0. The platform displacement prediction model 42 receives the initial arrays S1(1) and S2(1) to calculate multiple platform displacement fractions Q(1) (or Q(N)) and their respective corresponding relative displacements A(1) (or A(N)) in the first round (or the Nth round), as shown in step 502. Referring to step 503, the repair platform simulator 41 selects the higher of the platform displacement fractions Q(1) (or Q(N)) and uses the relative displacement A(1) (or A(N)) corresponding to the highest score to simulate the movement between the carrier substrate and the target substrate and complete the chip transfer. After the first (or Nth) round of transfer, the chip distribution on the carrier substrate and the target substrate has changed. The arrays representing the first position data and the second position data simulated by the repair platform simulator 41 are updated to S1(2) (or S1(N+1)) and S2(2) (or S2(N+1)) respectively, thereby generating the number of chips successfully transferred this time R(2) (or R(N+1)).

[0068] The platform displacement fraction Q(1) and the number of transferred chips R(2) are stored in the memory or storage device, as shown in step 504. The loss function model 43 uses the platform displacement fraction Q(1), the number R(2), and the platform displacement fraction Q(2) to calculate the loss value, minimize the loss value, and obtain the parameter P in the updated platform displacement prediction model 42. When the loss value is less than a threshold, the training and learning process ends, as shown in step 506. When the loss value is not less than the threshold, as shown in step 507, the N value is increased by 1, and the next round of chip transfer simulation is performed until the loss value is less than the threshold, at which point the training ends.

[0069] Similarly, the updated platform displacement prediction model 42 calculates the platform displacement fraction Q(N) and the corresponding optimal relative displacement A(N) for the current Nth round based on the arrays S1(N) and S2(N), as shown in step 502. In step 503, the repaired platform simulator 41 calculates S1(N+1) and S2(N+1) after the Nth round of transfer based on the optimal relative displacement A(N), and calculates the number of successfully transferred chips R(N+1). The loss function model 43 uses the platform displacement fraction Q(N) of the Nth round, the number of successfully transferred chips R(N+1), and the platform displacement fraction Q(N-1) of the previous round to calculate the loss value, minimizes the loss value, and obtains the updated platform displacement prediction model 42 with parameter P, as shown in step 505.

[0070] The formula for calculating the loss value is as follows:

[0071] Loss = [R(N) + γQ(N) - Q(N-1)] 2 γ = 0.95 (This formula and coefficients are for illustrative purposes only).

[0072] (This does not limit the scope of this application).

[0073] Figure 6 To depict the architecture of the platform displacement prediction model, virtual initial arrays S1 and S2, representing the first position data (chip on the carrier substrate) and the second position data (chip on the target substrate), are input into the machine learning model 421 in the platform displacement prediction model 42. Based on the parameter P 4211 that is updated in each round of the model, the platform displacement score Q for that round is calculated. The function block that retrieves the relative displacement 422 corresponding to the highest score is executed, and finally the optimal relative displacement A is output.

[0074] Figure 7To illustrate the flowchart of the repair process performed by the predictive machine model, as shown in step 701, during the Nth round of chip transfer (N=1 if it is the 1st round), the image recognition module 14 examines the chips on the carrier substrate and the target substrate, and outputs an array S1(N) representing the first position data of the chips on the carrier substrate and an array S2(N) representing the second position data of the chips on the target substrate. The predictive machine model uses the arrays S1(N) and S2(N) from the Nth round to calculate the optimal relative displacement A(N) for this round, as shown in step 702. In step 703, the transfer device performs the optimal relative displacement A(N) for the Nth round, calculates the arrays S1(N+1) and S2(N+2) representing the chips on the carrier substrate and the target substrate respectively after this round of transfer, and calculates the number of chip vacancies on the target substrate. If the number of chip vacancies is zero or less than a predetermined value, the chip repair or implantation is complete, and the repair operation on the target substrate ends, as shown in step 704. If the number of chip vacancies is not zero or is greater than a predetermined value, then proceed to step 705, increment the value of N by one, that is, perform the next round of repair work, repeat steps 702 to 703 until it is confirmed that the chip repair or implantation has been completed.

[0075] [Calculations for training a predictive machine learning model]

[0076] The array S1 of the first position data of the chip distribution on the carrier substrate is an H1×W1 array, and the initial array S2 of the second position data of the chip distribution on the target substrate is an H2×W2 array. Arrays S1 and S2 can be represented as follows:

[0077]

[0078] Here It can be 0 or 1; It can be 0 or 1; H1×W1=N1; H2×W2=N2.

[0079] See Figure 6 The machine learning model 421 contains a function with parameter P, whose independent variables are arrays S1 and S2. Substituting these values ​​yields multiple function values. Each value represents a platform displacement fraction Q corresponding to a relative displacement. If there are L possible relative displacements A, the function simultaneously outputs L platform displacement fractions Q, namely q1 to q2. L :

[0080] q1 = f1(S1, S2; P);

[0081] q2 = f2(S1, S2; P);

[0082] …

[0083] q L =f L(S1,S2;P);

[0084] Alternatively, the above expressions can be combined and expressed as:

[0085] Q = F(S1, S2; P);

[0086] Here, Q is q1 to q L The set of F; F is f1~f L The set; parameter P = {p i}, set all parameters p i The initial value is a random number.

[0087] Search platform displacement fractions Q of q1~q L The highest scorer is represented by MAX(F(S1,S2;P)), and their corresponding relative displacement A is the optimal relative displacement A*. The virtual carrier platform and virtual target platform in the platform simulator 41 will simulate motion based on this optimal relative displacement A*, thus obtaining the number of transferred chips R. Next, a new round of chip transfer will begin. The optimal relative displacement A* from the previous round will be renamed A′, and the arrays S1 and S2 from the previous round will be rewritten as S1′ and S2′ respectively. The platform displacement score Q from the previous round will then be changed to Q′.

[0088] Substituting the rewritten S1′, S2′, A′, and Q′ from the previous round, and the S1, S2, A*, and Q from this round, we can change the parameter p within parameter P. i Calculate and update using the following formula:

[0089]

[0090] Here, α = 0.05; γ = 0.95; if q A′ It was calculated using a neural network-like algorithm. It can be obtained using currently open-source neural network development platforms (such as Tensorflow, PyTorch, or Mxnet).

[0091] [Example of a predictive machine model generating optimal relative displacement]

[0092] Figure 8A and Figure 8B This is a schematic diagram depicting the chip distribution on a carrier substrate and a target substrate according to an embodiment of the present invention. The initial array S1(1) of the first position data of the chip distribution on the carrier substrate 81a and the initial array S2(1) of the second position data of the chip distribution on the target substrate 82b are respectively represented as follows:

[0093] S1=[x1x2]=[1 1], H1=1, W1=2, N1=2;

[0094] S2=[x3x4]=[0 0], H2=2, W2=1, N2=2.

[0095] And parameter p i =w mn ,1≤m≤4,1≤n≤4;p i =b l , 1≤l≤4. The two parameter values ​​are shown below:

[0096] [w 11 w 12 w 13 w 14 w 21 w 22 w 23 w 24 w 31 w 32 w 33 w 34 w 41 w 42 w 43 w 44 ]=

[0097] [0.323 -0.319 0.313 0.165-1.489 1.107-0.7408-

[0098] 0.526 0.648 0.245-1.661 -0.526 2.761 0.641 0.294 -2.469]; and

[0099] [b1b2b3b4] = [0 0 0 0].

[0100] The platform displacement fraction Q can be calculated using the following formula:

[0101] Q = [q1q2q3q4]

[0102] =[w 11 w 12 w 13 w 14 w 21 w 22 w 23 w 24 w 31 w 32 w 33 w 34 w 41 w 42 w 43 w 44 ]

[0103] [x1x2x3x4]+[b1b2b3b4]=[0.004 -0.382 0.893 3.402].

[0104] See Figures 8C to 8F This diagram illustrates four possible relative displacements between the carrier substrate and the target substrate. q1 corresponds to... Figure 8C The first A1-1 in the set of relative displacements A(1) is (-1, 0); q2 corresponds to Figure 8D The second A1-2 in the set of relative displacements A(1) is (0, 0); q3 corresponds to Figure 8E The third A1-3 in the set of relative displacements A(1) is (-1, -1); q4 corresponds to Figure 8F The fourth A1-4 in the set of relative displacements A(1) is (0, -1).

[0105] The maximum value in Q = [0.004 - 0.382 0.893 3.402] is q4 = 3.402, which corresponds to the 4th relative displacement A1-4. Therefore, the optimal relative displacement A* = (0, -1). To simulate the next round of chip transfer, let A′ = (0, -1), S1′ =

[11] , S2′ =

[00] , Q′ = [0.004 - 0.382 0.893 3.402]. At the start of the second round of chip transfer simulation, arrays S1(2) and S2(2) are represented as S1 =

[01] and S2 =

[01] . Therefore, the number of successfully transferred chips R = 1.

[0106] Recalculate the platform displacement fraction for the second round, and we get Q = [0.323 -0.319 0.313 0.165 -1.489 1.107 -0.7408 -0.526 0.648 0.245 -1.661 -0.526 2.761 0.641 0.294 -2.469][0 1 0 1] + [0 0 0 0] = [0.155 2.957 -0.281 -1.828]. After the update, the maximum value of Q = [0.155 2.957-0.281 -1.828] is q2 = 2.957, which corresponds to the second displacement. This corresponds to the second relative displacement A1-2. Therefore, the optimal relative displacement A* = (0, 0) is obtained. Thus, the carrier platform and the target platform complete the relative motion and subsequent chip transfer with the optimal relative displacement A* = (0, 0).

[0107] Parameter P = {p i Updated to:

[0108]

[0109] q A*(S1, S2; P) = q2 (S1, S2; P) = 2.957;

[0110] q A′ (S′1, S′2; P) = q4 (S1′, S2′; P) = 3.402;

[0111] α=0.05, γ=0.95, R=1.

[0112] And when parameter p i =w 4n Then the original parameter w 4n =[2.761 0.641 0.294 -2.469], which will be updated according to the aforementioned calculation formula:

[0113] [2.761 0.641 0.294 -2.469]+0.05×(1+0.95×2.957-3.402)×[1 1 0 0]=[2.781 0.661 0.294 -2.469].

[0114] Furthermore, when parameter p i =b4, then Therefore, the original b4 = 0 will be updated to b4 = 0 + 0.05 × (1 + 0.95 × 2.957 - 3.402) × 1 = 0.02.

[0115] For other parameters p i =w mn and b m ,because Therefore, no update is made. Thus, the updated parameter P = {p i}={w mn b m} represents the values ​​in the following arrays:

[0116] w mn =[0.323 -0.319 0.313 0.165 -1.489 1.107 -0.7408 -

[0117] 0.526 0.648 0.245 -1.661 -0.526 2.781 0.661 0.294 -2.46] and b m =

[0118] [0 0 0 0.02].

[0119] Since the array S2, representing the second position data of the chip distribution on the target substrate, is currently

[01] , the aforementioned calculation of the optimal relative displacement must be performed again to complete the repair of the last gap. That is, when S2 =

[11] , it indicates that the chip repair or implantation has been completed. The aforementioned formulas and coefficients are merely illustrative and do not limit the scope of this application.

[0120] Although this invention has been written with reference to specific embodiments and implementations, various changes and modifications will be apparent to those skilled in the art. The aim is to include such changes and modifications that fall within the scope of the appended patent application.

Claims

1. A system for repairing defects on a substrate, characterized in that, The method comprises: a transfer device configured to transfer at least one first die on a first substrate to at least one vacancy on a second substrate; an image recognition module configured to respectively inspect and obtain a first position data of the at least one first die on the first substrate and a second position data of the at least one second die on the second substrate; and an operation processing module configured to input the first position data and the second position data into a prediction machine model to obtain an optimal relative displacement; wherein the transfer device generates a movement between the second substrate and the first substrate with the optimal relative displacement to perform the transfer of the at least one first die.

2. The system for filling a void on a substrate of claim 1, wherein, The optimal relative displacement is obtained by adjusting parameters in the prediction machine model to reduce a loss value.

3. The system for filling a void on a substrate of claim 1, wherein, After the transfer device completes the transfer of the at least one first die, the number of vacancies still existing on the second substrate is calculated.

4. The system for filling a void on a substrate of claim 3, wherein, If the number of vacancies still existing is not 0, the first position data and the second position data are updated, and another optimal relative displacement required for the current next round of transfer of dies is calculated.

5. The system for filling a void on a substrate of claim 1, wherein, The formation of the prediction machine model comprises: a repair platform simulator configured to generate a virtual first position data and a virtual second position data; a platform displacement prediction model configured to input the virtual first position data and the virtual second position data to calculate a plurality of platform displacement scores, select a best virtual relative displacement corresponding to a highest score from the platform displacement scores, and provide the best virtual relative displacement to the repair platform simulator to generate a simulated movement and a simulated transfer of dies, and output a transfer number; and a loss function model configured to update parameters in the platform displacement prediction model according to the transfer number and the platform displacement scores.

6. A method of repairing a void in a substrate, the method comprising: The method comprises: inspecting at least one first die on a first substrate to obtain a first position data and inspecting at least one second die on a second substrate to obtain a second position data; inputting the first position data and the second position data into a prediction machine model to obtain an optimal relative displacement; generating a movement between the second substrate and the first substrate with the optimal relative displacement; and transferring the at least one first die to the at least one vacancy.

7. The method of claim 6, wherein the method further comprises: The step of obtaining the optimal relative displacement is performed by adjusting parameters in the prediction machine model to reduce a loss value, thereby obtaining the optimal relative displacement.

8. The method of claim 6, wherein the method further comprises: The method further comprises, after completing the transfer of the at least one first die, calculating the number of vacancies still existing on the second substrate.

9. The method of claim 8, wherein the method further comprises: If the number of vacancies still existing is not 0, the first position data and the second position data are updated, and another optimal relative displacement required for the current next round of transfer of dies is calculated.

10. The method of claim 6, wherein the method further comprises: The formation of the prediction machine model is performed by the following steps: generating a virtual first position data and a virtual second position data; inputting the virtual first position data and the virtual second position data to calculate a plurality of platform displacement scores; selecting a best virtual relative displacement corresponding to a highest score from the platform displacement scores; simulating a movement between a virtual first substrate and a virtual second substrate and simulating a transfer of dies with the best virtual relative displacement, and outputting a transfer number; and The parameters in the predictive machine model are updated based on the number of transfers and the displacement fractions of these platforms.

Images