A silicon wafer grinding profile prediction and compensation method, device, equipment and medium

Abstract

The application relates to a silicon wafer grinding profile prediction and compensation method, device, equipment and medium, wherein the silicon wafer grinding profile prediction and compensation method comprises the following steps: obtaining training samples based on process parameters obtained in a historical grinding process of a silicon wafer and thickness and profile parameters of the silicon wafer measured after grinding; obtaining a silicon wafer profile prediction model by training the training samples based on a neural network; inputting current process parameters obtained in a grinding process of the silicon wafer into the silicon wafer profile prediction model to obtain thickness and profile parameter prediction values of the silicon wafer; and adjusting the current process parameters based on the thickness and profile parameter prediction values and thickness and profile parameter set values meeting process standards, so that the thickness and profile parameter prediction values meet the thickness parameter set value of the silicon wafer. The automatic compensation of the thickness parameter set value of the silicon wafer meeting the process standards is realized, and the production efficiency, precision and consistency of the finished silicon wafer are improved.

Description

Technical Field

[0001] This application relates to the field of semiconductor silicon wafer processing technology, and in particular to a method, apparatus, equipment and medium for predicting and compensating for the grinding morphology of silicon wafers. Background Technology

[0002] As integrated circuits have evolved from small-scale to ultra-large-scale production, the diameter of single-crystal silicon wafers used in integrated circuits is continuously increasing, thus raising the requirements for the surface morphology precision of silicon wafers. Silicon wafer thinning, a precision grinding step in silicon wafer manufacturing, can remove thickness and improve the total thickness variation (TTV) parameter of the silicon wafer before polishing, increasing the efficiency of the polishing process. It can also improve the surface grinding quality of the silicon wafer, removing surface damage caused by the grinding process. Therefore, after grinding, the silicon wafer thinning machine must ensure that the silicon wafer has a stable average thickness (THK), a stable total thickness variation (TTV), and excellent surface quality. The silicon wafer thinning machine allows adjustment of the silicon wafer grinding morphology and compensation of the grinding thickness via an operation panel during the grinding process.

[0003] Currently, the parameters (THK, TTV) of silicon wafers processed by different grinding processes in silicon wafer thinning machines rely on the measurement of the measuring equipment. During the silicon wafer production process, process personnel need to manually adjust and compensate based on the sampled silicon wafer parameters and their personal experience. This reduces production efficiency to some extent and affects the accuracy and consistency of silicon wafer parameters. There is a lack of automatic targeted identification and analysis methods for grinding processes and processed silicon wafer parameters, as well as automatic compensation methods for processed silicon wafer parameters. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, apparatus, equipment, and medium for predicting and compensating the grinding morphology of silicon wafers to address the aforementioned technical problems.

[0005] In a first aspect, embodiments of this application provide a method for predicting and compensating for the grinding morphology of silicon wafers, the method comprising:

[0006] Training samples were obtained based on the process parameters affecting grinding quality obtained during the historical grinding process of silicon wafers and the thickness and morphology parameters of silicon wafers measured after grinding.

[0007] A silicon wafer morphology prediction model is obtained by training the training samples based on a neural network.

[0008] The current process parameters obtained during the current grinding process of the silicon wafer are input into the silicon wafer morphology prediction model to obtain the predicted values ​​of silicon wafer thickness and morphology parameters.

[0009] Based on the predicted values ​​of silicon wafer thickness and morphology parameters and the set values ​​of silicon wafer thickness and morphology parameters that meet the process standards, the current process parameters are adjusted so that the predicted values ​​of silicon wafer thickness and morphology parameters meet the set values ​​of silicon wafer thickness and morphology parameters.

[0010] In one embodiment, the process parameters include at least one of the following: grinding process parameters, grinding disc tilt angle parameters, grinding temperature parameters, grinding water flow rate parameters, and sensor thickness measurement parameters.

[0011] In one embodiment, the grinding process parameters include spindle speed, grinding disc speed, spindle feed speed, and grinding time; the grinding temperature parameters include spindle cooling water temperature and grinding pure water temperature; the grinding water flow parameters include spindle center water flow, water spray flow in the water tank, water dripping flow at the thickness measuring rod, and cooling water flow at the thickness measuring head; the silicon wafer thickness and morphology parameters include the average silicon wafer thickness measured by the measuring machine and the total silicon wafer thickness deviation; the sensor thickness measurement parameters include the average silicon wafer thickness, center thickness, maximum thickness, minimum thickness, initial thickness, and total silicon wafer thickness deviation measured by the sensor.

[0012] In one embodiment, the neural network is a backpropagation (BP) neural network, comprising an input layer, a hidden layer, and an output layer; the step of training the training samples based on the neural network to obtain a silicon wafer morphology prediction model includes:

[0013] The process parameters are used as input parameters, and the silicon wafer thickness and morphology parameters are used as output parameters.

[0014] Based on the input parameters and the output parameters, the number of neurons in the input layer, the hidden layer, and the output layer is determined to construct a BP neural network;

[0015] Based on the input parameters and the output parameters, a silicon wafer morphology prediction model is obtained by training the training samples using a constructed BP neural network.

[0016] In one embodiment, the step of training the training samples using a constructed BP neural network to obtain a silicon wafer morphology prediction model based on the input parameters and the output parameters includes:

[0017] The input parameters are used as input variables of the input layer of the BP neural network. During the forward propagation process, the input variables undergo nonlinear transformation through the hidden layer to obtain the output variable value, and the output variable value is output by the output layer.

[0018] Calculate the error between the output variable value and the theoretically set output parameters;

[0019] If the error value is greater than the expected error, the weight and threshold of each neuron in the BP neural network are adjusted through the backpropagation algorithm until the error value is less than the expected error, thereby obtaining the silicon wafer morphology prediction model.

[0020] In one embodiment, adjusting the current process parameters based on the predicted values ​​of the silicon wafer thickness and morphology parameters and the set values ​​of the silicon wafer thickness and morphology parameters conforming to the process standards, so that the predicted values ​​of the silicon wafer thickness and morphology parameters meet the set values ​​of the silicon wafer thickness and morphology parameters, includes:

[0021] If the predicted value of the average thickness of the silicon wafer does not meet the set value of the average thickness of the silicon wafer, the initial thickness parameter in the current sensor thickness measurement parameters is adjusted.

[0022] If the predicted value of the total thickness deviation of the silicon wafer does not meet the set value of the total thickness deviation of the silicon wafer, the grinding disc tilt angle parameter in the current process parameters will be adjusted.

[0023] In one embodiment, the method further includes: after the current round of training of the BP neural network ends, if the error value is greater than the expected error, then update the training steps of the BP neural network and start a new round of training.

[0024] Secondly, embodiments of this application also provide a silicon wafer grinding morphology prediction and compensation device, the device comprising:

[0025] The sample acquisition module is used to obtain training samples based on the process parameters affecting grinding quality obtained during the historical grinding process of silicon wafers and the thickness and morphology parameters of silicon wafers measured after grinding.

[0026] The training module is used to train the training samples based on a neural network to obtain a silicon wafer morphology prediction model;

[0027] The prediction module is used to input the current process parameters obtained during the current grinding process of the silicon wafer into the silicon wafer morphology prediction model to obtain the predicted values ​​of silicon wafer thickness and morphology parameters.

[0028] The adjustment module is used to adjust the current process parameters based on the predicted values ​​of the silicon wafer thickness and morphology parameters and the set values ​​of the silicon wafer thickness and morphology parameters that meet the process standards, so that the predicted values ​​of the silicon wafer thickness and morphology parameters meet the set values ​​of the silicon wafer thickness and morphology parameters.

[0029] Thirdly, embodiments of this application also provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the method described in the first aspect above.

[0030] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the method described in the first aspect above.

[0031] The aforementioned silicon wafer grinding morphology prediction and compensation method, apparatus, equipment, and medium obtain training samples based on process parameters affecting grinding quality acquired during historical grinding processes and silicon wafer thickness and morphology parameters measured after grinding. A silicon wafer morphology prediction model is obtained by training these training samples using a neural network. Current process parameters acquired during the current grinding process are input into the silicon wafer morphology prediction model to obtain predicted values ​​for silicon wafer thickness and morphology parameters. Based on these predicted values ​​and the set values ​​conforming to process standards, the current process parameters are adjusted to ensure that the predicted values ​​meet the set values. This achieves automatic parameter compensation for silicon wafer thickness and morphology parameters conforming to process standards, improving the production efficiency, accuracy, and consistency of finished silicon wafers.

[0032] Details of one or more embodiments of this application are set forth in the following drawings and description to make other features, objects and advantages of this application more readily apparent. Attached Figure Description

[0033] Figure 1 This is a hardware structure block diagram of a terminal device for a silicon wafer grinding morphology prediction and compensation method in one embodiment;

[0034] Figure 2 This is a flowchart illustrating a method for predicting and compensating for silicon wafer grinding morphology in one embodiment;

[0035] Figure 3 This is a schematic diagram of the logic flow of the BP neural network algorithm in one embodiment;

[0036] Figure 4 This is a flowchart illustrating the silicon wafer grinding morphology prediction and compensation method in another embodiment;

[0037] Figure 5 This is a structural block diagram of a silicon wafer grinding morphology prediction and compensation device in one embodiment;

[0038] Figure 6 This is a schematic diagram of the computer device structure in one embodiment. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of this application clearer, the application is described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application.

[0040] Obviously, the accompanying drawings described below are merely some examples or embodiments of this application. Those skilled in the art can apply this application to other similar scenarios based on these drawings without any creative effort. Furthermore, it is understood that although the efforts made in this development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, any changes to design, manufacturing, or production based on the technical content disclosed in this application are merely conventional technical means and should not be construed as insufficient disclosure of this application.

[0041] The method embodiments provided in this example can be executed on a terminal, computer, or similar computing device. For example, it can run on a terminal. Figure 1 This is a hardware structure block diagram of the terminal of the silicon wafer grinding morphology prediction and compensation method in this embodiment. For example... Figure 1 As shown, a terminal may include one or more ( Figure 1 (Only one is shown) A processor 102 and a memory 104 for storing data. The processor 102 may be, but is not limited to, a processing device such as a microprocessor (MCU) or a programmable logic device (FPGA). The terminal may also include a transmission device 106 for communication functions and an input / output device 108. The input device is used to input the process parameters of the silicon wafer during the grinding process (including grinding process parameters (spindle speed N1, grinding disc speed N2, spindle feed speed V1, grinding time t1), grinding disc tilt parameters (A1, A2, B1, B2), temperature parameters during grinding (spindle cooling water temperature T1, grinding pure water temperature T2), grinding water flow parameters (spindle center water flow Q1, water spray flow in water tank Q2, water dripping flow at thickness measuring rod Q3, cooling water flow at thickness measuring head Q4), and sensor thickness measurement parameters (average silicon wafer thickness Z1, silicon wafer center thickness Z2, silicon wafer maximum thickness Z3, silicon wafer minimum thickness Z4, silicon wafer initial thickness Z5, silicon wafer total thickness deviation Z6). The output device is used to output the silicon wafer thickness and morphology parameters.

[0042] Those skilled in the art will understand that Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the terminal described above. For example, the terminal may also include components that are larger than... Figure 1The more or fewer components shown, or having the same Figure 1 The different configurations shown are illustrated.

[0043] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the silicon wafer grinding morphology prediction and compensation method in this embodiment. The processor 102 executes various functional applications and data processing by running the computer programs stored in the memory 104, thereby implementing the above-described method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0044] The transmission device 106 is used to receive or send data via a network. This network includes a wireless network provided by the terminal's communication provider. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 can be a Radio Frequency (RF) module for wireless communication with the Internet.

[0045] This application provides a method for predicting and compensating the grinding morphology of silicon wafers, which is applied to... Figure 1 Taking terminal 102 as an example, the explanation is as follows: Figure 2 As shown, the method includes the following steps:

[0046] S201, training samples are obtained based on the process parameters affecting grinding quality obtained during the historical grinding process of silicon wafers and the thickness and morphology parameters of silicon wafers measured after grinding.

[0047] In this embodiment, different process parameters are used during the grinding of silicon wafers by a thinning machine, and the silicon wafer thickness and morphology parameters are measured by a measuring machine after grinding to obtain multiple training samples. The silicon wafer thickness parameter is the average thickness THK of the silicon wafer, and the silicon wafer morphology parameter is the total thickness deviation TTV of the silicon wafer.

[0048] S202, The training samples are trained based on a neural network to obtain a silicon wafer morphology prediction model;

[0049] S203, input the current process parameters obtained during the current grinding process of the silicon wafer into the silicon wafer morphology prediction model to obtain the predicted values ​​of silicon wafer thickness and morphology parameters;

[0050] S204, based on the predicted values ​​of the silicon wafer thickness and morphology parameters and the set values ​​of the silicon wafer thickness and morphology parameters that meet the process standards, the current process parameters are adjusted so that the predicted values ​​of the silicon wafer thickness and morphology parameters meet the set values ​​of the silicon wafer thickness and morphology parameters.

[0051] The silicon wafer grinding morphology prediction and compensation method provided in this embodiment obtains training samples based on process parameters affecting grinding quality acquired during historical grinding processes and silicon wafer thickness and morphology parameters measured after grinding. A silicon wafer morphology prediction model is obtained by training the training samples using a neural network. Current process parameters acquired during the current grinding process are input into the silicon wafer morphology prediction model to obtain predicted values ​​for silicon wafer thickness and morphology parameters. Based on the predicted values ​​and the set values ​​conforming to process standards, the current process parameters are adjusted to ensure that the predicted values ​​meet the set values. This provides an automatic targeted identification and analysis method for process parameters and silicon wafer processing parameters, achieving automatic compensation for silicon wafer thickness and morphology parameter settings that conform to process standards. This improves the production efficiency, accuracy, and consistency of finished silicon wafers.

[0052] In one embodiment, the process parameters include at least one of the following: grinding process parameters, grinding disc tilt angle parameters, grinding temperature parameters, grinding water flow rate parameters, and sensor thickness measurement parameters.

[0053] The morphology of silicon wafers ground by a thinning machine is mainly determined by grinding parameters such as grinding process parameters and grinding disc tilt angle. Under fixed grinding process and grinding disc tilt angle, the morphology of silicon wafers is also affected by factors such as pure water temperature, cooling water temperature, grinding water flow rate, and thickness measurement water flow rate during the grinding process.

[0054] In one embodiment, the grinding process parameters include spindle speed N1, grinding disc speed N2, spindle feed speed V1, and grinding time t1; grinding disc tilt parameters (A1, A2, B1, B2); grinding temperature parameters include spindle cooling water temperature T1 and grinding pure water temperature T2; grinding water flow parameters include spindle center water flow Q1, water spray flow in the water tank Q2, drip flow at the thickness measuring rod Q3, and cooling water flow at the thickness measuring head Q4; the sensor thickness measurement parameters include the average silicon wafer thickness Z1, silicon wafer center thickness Z2, maximum silicon wafer thickness Z3, minimum silicon wafer thickness Z4, initial silicon wafer thickness Z5, and total silicon wafer thickness deviation Z6 measured by the thickness measuring sensor. The silicon wafer thickness and morphology parameters include the average silicon wafer thickness THK and total silicon wafer thickness deviation TTV measured by the measuring machine.

[0055] In one embodiment, the neural network is a backpropagation (BP) neural network, comprising an input layer, a hidden layer, and an output layer; the step of training the training samples based on the neural network to obtain a silicon wafer morphology prediction model includes the following:

[0056] The process parameters are used as input parameters, and the silicon wafer thickness and morphology parameters are used as output parameters. Based on the input parameters and the output parameters, the number of neurons in the input layer, the hidden layer, and the output layer are determined to construct a BP neural network. Based on the input parameters and the output parameters, the constructed BP neural network is trained to obtain a silicon wafer morphology prediction model.

[0057] In a specific embodiment, the process parameters affecting grinding quality are obtained as follows: grinding process parameters (spindle speed N1, grinding disc speed N2, spindle feed speed V1, polishing time t1), grinding disc tilt angle parameters (A1, A2, B1, B2), temperature parameters during grinding (spindle cooling water temperature T1, grinding pure water temperature T2), grinding water flow rate parameters (spindle center water flow rate Q1, water spray flow rate in water tank Q2, water dripping flow rate at thickness measuring rod Q3, cooling water flow rate at thickness measuring head Q4), and sensor thickness measurement parameters (average silicon wafer thickness Z1, silicon wafer center thickness Z2, maximum silicon wafer thickness Z3, minimum silicon wafer thickness Z4, initial silicon wafer thickness Z5, total silicon wafer thickness deviation Z6). The above 20 process parameters are used as input parameters x={N1,N2,V1,t1,A1,A2,B1,B2,T1,T2,Q1,Q2,Q3,Q4,Z1,Z2,Z3,Z4,Z5,Z6} of the BP neural network. Therefore, the number of neurons in the input layer of the BP neural network can be determined to be m=20.

[0058] In this embodiment, the silicon wafer thickness and morphology parameters are specifically the average silicon wafer thickness THK and the total silicon wafer thickness deviation TTV measured by the measuring instrument. The above two silicon wafer thickness and morphology parameters are used as the output parameters of the BP neural network. The output parameters can be expressed as o={THK,TTV}. Therefore, the number of neurons in the output layer of the BP neural network can be determined to be l=2.

[0059] In this embodiment, the number q of hidden layer neurons in the BP neural network is determined by... Determine, where m is the number of neurons in the input layer, and a is an empirical value, ranging from 1 to 10. For example, the empirical value a can be set to 1, and then the number of neurons in the hidden layer can be calculated. A backpropagation (BP) neural network is constructed based on the number of neurons in the input layer, hidden layer, and output layer. The BP neural network is then trained based on the input and output parameters. After training, a silicon wafer morphology prediction model is obtained.

[0060] In one embodiment, the step of training the constructed BP neural network based on the input parameters and the output parameters to obtain the silicon wafer morphology prediction model includes the following:

[0061] The input parameters are used as input variables for the input layer of a BP neural network. During forward propagation, the input variables undergo nonlinear transformation through the hidden layer to obtain output variable values, which are then output by the output layer. The error between the output variable values ​​and the theoretically set output parameters is calculated. If the error value is greater than the expected error, the weights and thresholds of each neuron in the BP neural network are adjusted using an error backpropagation algorithm until the error value is less than the expected error, thus obtaining a silicon wafer morphology prediction model.

[0062] Backpropagation (BP) neural networks can correct nonlinear systems with complex input-output relationships and eliminate the influence of interference factors. Training a BP neural network mainly involves two processes: the forward propagation of the input signal and the reverse error control process. During forward propagation, the input signal undergoes an internal nonlinear transformation to output a signal. If the error between the output signal value and the ideal value exceeds a set range, the BP neural network enters the reverse error control process, feeding the error back to the internal neurons. By adjusting the weights and thresholds, the output error is reduced to a specified value, at which point training ends. Table 1 shows the training parameters for the BP neural network.

[0063] Table 1

[0064]

[0065] The forward propagation process of a BP neural network is as follows:

[0066] The input to the i-th node of the hidden layer is:

[0067]

[0068] Where, ω ij θ is the weighting coefficient. i The threshold value is used.

[0069] The output of the i-th node in the hidden layer is:

[0070]

[0071] The hidden layer activation function is:

[0072]

[0073] The input to the k-th node of the output layer is:

[0074]

[0075] Where, ω ki α is the weighting coefficient. k The threshold value is used.

[0076] The output of the k-th node in the output layer is:

[0077]

[0078] The activation function of the output layer is:

[0079] ψ(net k ) = pureline(net k ) = net k

[0080] This application's embodiments, combining the input and output parameters of the silicon wafer morphology prediction model, describe the forward propagation process of the BP neural network:

[0081] Hidden layer input matrix NET1:

[0082] NET1=[net1 net2 net3 net4 net5 net6] T =ω1×x+θ

[0083]

[0084] x=[N1 N2 V1 t1 A1 A2 B1 B2 T1 T2 Q1 Q2 Q3 Q4 Z1 Z2] T

[0085] θ=[θ1 θ2 θ3 θ4 θ5 θ6] T

[0086] Hidden layer output matrix y:

[0087] y = [y1 y2 y3 y4 y5 y6] T

[0088]

[0089] Output layer input matrix NET2:

[0090] NET 2 = [net 1 net 2 ] T =ω2×y+α

[0091]

[0092] y = [y1 y2 y3 y4 y5 y6] T

[0093] α=[α1 α2] T

[0094] Output layer output matrix o:

[0095] o = [o1 o2] T =[net 1 net 2 ] T

[0096] The backpropagation process of the BP neural network is as follows:

[0097] Based on the error between the output and the theoretical setting, the weights and thresholds of each neuron in the network are adjusted using the error backpropagation algorithm to reduce the error value and improve the accuracy of the network.

[0098] Assuming the total number of samples is P, then the quadratic error criterion function is E. P :

[0099]

[0100] Among them, T k For the theoretical output parameters THK, TTV, o k The output variable values ​​are THK and TTV for the BP neural network.

[0101] The total error criterion function is:

[0102]

[0103] The adjustments made to the weights and thresholds during backpropagation are as follows:

[0104]

[0105]

[0106]

[0107] Where η is the learning rate.

[0108] The adjusted weights and thresholds are as follows:

[0109]

[0110]

[0111]

[0112] Based on the above settings, a BP neural network is trained using the input parameters to obtain a silicon wafer morphology prediction model. This embodiment provides an automatic targeted identification and analysis method for process parameters and silicon wafer processing parameters, realizing automatic compensation of silicon wafer processing parameters THK and TTV, improving the production efficiency of finished silicon wafers, and enhancing the accuracy and consistency of finished silicon wafers.

[0113] like Figure 3 As shown, to more clearly illustrate the training process of the BP neural network in this embodiment, this embodiment further demonstrates the algorithm logic of the BP neural network, including the following steps:

[0114] S301: Initialize the weights and thresholds of the BP neural network;

[0115] S302: Input training samples;

[0116] S303: Calculate the input and output of each neuron in the hidden layer;

[0117] S304: Calculate the input and output of each neuron in the output layer;

[0118] S305: Calculate the output layer error;

[0119] S306: Calculate the hidden layer error;

[0120] S307: Correct the weights and thresholds of the output layer and hidden layer;

[0121] S308: Determine whether all training sample data has been trained; if yes, proceed to S309; ​​if no, start a new training step.

[0122] S309: Determine if the output layer error is less than the preset value e; if yes, the current training round ends; if no, execute S310.

[0123] S310: Update training steps and start a new round of training.

[0124] In one embodiment, adjusting the current process parameters based on the predicted values ​​of the silicon wafer thickness and morphology parameters and the set values ​​of the silicon wafer thickness and morphology parameters conforming to the process standards, so that the predicted values ​​of the silicon wafer thickness and morphology parameters meet the set values ​​of the silicon wafer thickness and morphology parameters, includes the following:

[0125] If the predicted value of the average thickness of the silicon wafer does not meet the set value of the average thickness of the silicon wafer, the initial thickness parameter in the current sensor thickness measurement parameters is adjusted; if the predicted value of the total thickness deviation of the silicon wafer does not meet the set value of the total thickness deviation of the silicon wafer, the grinding disc tilt angle parameter in the current process parameters is adjusted.

[0126] Specifically, when the predicted value of the average thickness of the silicon wafer does not meet the ideal set value of the average thickness of the silicon wafer, the initial thickness parameter in the thickness measurement parameters of the sensor can be adjusted first, or other process parameters can be adjusted, such as adjusting the grinding process parameters (spindle speed N1, grinding disc speed N2, spindle feed speed V1, polishing time t1); the total thickness deviation of the silicon wafer is mainly affected by the grinding disc tilt angle parameters (A1, A2, B1, B2). Therefore, when the predicted value of the total thickness deviation of the silicon wafer does not meet the ideal set value of the total thickness deviation of the silicon wafer, the grinding disc tilt angle parameter can be adjusted first.

[0127] In one embodiment, the method further includes: after the current round of training of the BP neural network ends, if the error value is greater than the expected error, then update the training steps of the BP neural network and start a new round of training.

[0128] like Figure 4 As shown, in one embodiment, a method for predicting and compensating the grinding morphology of a silicon wafer is provided, the method comprising the following steps:

[0129] S401: Input parameters are acquired during the silicon wafer grinding process, including grinding process parameters, grinding disc tilt angle parameters, grinding temperature parameters, grinding water flow rate parameters, and sensor thickness measurement parameters.

[0130] S402: After the wafer grinding is completed, measure the silicon wafer to obtain output parameters, including the average thickness of the silicon wafer THK and the total thickness deviation of the silicon wafer TTV;

[0131] S403: The input and output parameter sample data are trained using a BP neural network to obtain a silicon wafer morphology prediction model.

[0132] S404: Collects input parameters in real time during the grinding process, obtains the ideal grinding process, tilt angle parameters, and thickness measurement parameters through the prediction algorithm model, and finally adjusts the thinning machine grinding process parameters, tilt angle parameters, and thickness sensor parameters to the ideal parameter values ​​in real time.

[0133] This application also provides a silicon wafer grinding morphology prediction and compensation device, such as... Figure 5 As shown, the device includes: a sample acquisition module 10, a training module 20, a prediction module 30, and an adjustment module 40, wherein:

[0134] The sample acquisition module 10 is used to obtain training samples based on the process parameters affecting grinding quality obtained during the historical grinding process of the silicon wafer and the thickness and morphology parameters of the silicon wafer measured after grinding.

[0135] Training module 20 is used to train the training samples based on a neural network to obtain a silicon wafer morphology prediction model;

[0136] The prediction module 30 is used to input the current process parameters obtained during the current grinding process of the silicon wafer into the silicon wafer morphology prediction model to obtain the predicted values ​​of silicon wafer thickness and morphology parameters.

[0137] The adjustment module 40 is used to adjust the current process parameters based on the predicted values ​​of the silicon wafer thickness and morphology parameters and the set values ​​of the silicon wafer thickness and morphology parameters that meet the process standards, so that the predicted values ​​of the silicon wafer thickness and morphology parameters meet the set values ​​of the silicon wafer thickness and morphology parameters.

[0138] In one embodiment, the process parameters include at least one of the following: grinding process parameters, grinding disc tilt angle parameters, grinding temperature parameters, grinding water flow rate parameters, and sensor thickness measurement parameters.

[0139] In one embodiment, the grinding process parameters include spindle speed, grinding disc speed, spindle feed speed, and polishing time; the grinding temperature parameters include spindle cooling water temperature and grinding pure water temperature; the grinding water flow parameters include spindle center water flow, water spray flow in the water tank, water dripping flow at the thickness measuring rod, and cooling water flow at the thickness measuring head; the silicon wafer thickness and morphology parameters include the average silicon wafer thickness measured by the measuring machine and the total silicon wafer thickness deviation; the sensor thickness measurement parameters include the average silicon wafer thickness, center thickness, maximum thickness, minimum thickness, initial thickness, and total silicon wafer thickness deviation measured by the sensor.

[0140] In one embodiment, the neural network is a backpropagation (BP) neural network, comprising an input layer, a hidden layer, and an output layer; the training module 20 is further configured to:

[0141] The process parameters are used as input parameters, and the silicon wafer thickness and morphology parameters are used as output parameters.

[0142] Based on the input parameters and the output parameters, the number of neurons in the input layer, the hidden layer, and the output layer is determined to construct a BP neural network;

[0143] Based on the input parameters and the output parameters, a silicon wafer morphology prediction model is obtained by training the training samples using a constructed BP neural network.

[0144] In one embodiment, the training module 20 is further configured to:

[0145] The input parameters are used as input variables of the input layer of the BP neural network. During the forward propagation process, the input variables undergo nonlinear transformation through the hidden layer to obtain the output variable value, and the output variable value is output by the output layer.

[0146] Calculate the error between the output variable value and the theoretically set output parameters;

[0147] If the error value is greater than the expected error, the weight and threshold of each neuron in the BP neural network are adjusted through the backpropagation algorithm until the error value is less than the expected error, thereby obtaining the silicon wafer morphology prediction model.

[0148] In one embodiment, the adjustment module 40 is further configured to:

[0149] If the predicted value of the average thickness of the silicon wafer does not meet the set value of the average thickness of the silicon wafer, the initial thickness parameter in the current sensor thickness measurement parameters is adjusted.

[0150] If the predicted value of the total thickness deviation of the silicon wafer does not meet the set value of the total thickness deviation of the silicon wafer, the grinding disc tilt angle parameter in the current process parameters will be adjusted.

[0151] In one embodiment, the training module 20 is further configured to: after the current round of training of the BP neural network ends, if the error value is greater than the expected error, update the training steps of the BP neural network and start a new round of training.

[0152] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can reside in the same processor; or the above modules can be located in different processors in any combination.

[0153] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 6As shown, the computer device includes a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements a method for predicting and compensating the grinding morphology of silicon wafers. The display screen can be an LCD screen or an e-ink display screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.

[0154] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0155] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in any of the above embodiments of the silicon wafer grinding morphology prediction and compensation method.

[0156] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0157] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0158] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for predicting and compensating for the grinding morphology of silicon wafers, characterized in that, Applied to a thinning machine, the method includes: Training samples were obtained based on the process parameters affecting grinding quality obtained during the historical grinding process of silicon wafers and the thickness and morphology parameters of silicon wafers measured after grinding. The process parameters include: grinding process parameters, grinding disc tilt angle parameters, grinding temperature parameters, grinding water flow rate parameters, and sensor thickness measurement parameters. A silicon wafer morphology prediction model is obtained by training the training samples based on a neural network. The current process parameters obtained during the current grinding process of the silicon wafer are input into the silicon wafer morphology prediction model to obtain the predicted values ​​of silicon wafer thickness and morphology parameters. Based on the predicted values ​​of silicon wafer thickness and morphology parameters and the set values ​​of silicon wafer thickness and morphology parameters that meet the process standards, the current process parameters are adjusted so that the predicted values ​​of silicon wafer thickness and morphology parameters meet the set values ​​of silicon wafer thickness and morphology parameters.

2. The method according to claim 1, characterized in that, The grinding process parameters include spindle speed, grinding disc speed, spindle feed speed, and grinding time; the grinding temperature parameters include spindle cooling water temperature and grinding pure water temperature; the grinding water flow parameters include spindle center water flow, water spray flow in the water tank, water dripping flow at the thickness measuring rod, and cooling water flow at the thickness measuring head; the silicon wafer thickness and morphology parameters include the average silicon wafer thickness measured by the measuring machine and the total silicon wafer thickness deviation; the sensor thickness measurement parameters include the average silicon wafer thickness, center thickness, maximum thickness, minimum thickness, initial thickness, and total silicon wafer thickness deviation measured by the sensor.

3. The method according to claim 1, characterized in that, The neural network is a backpropagation (BP) neural network, comprising an input layer, hidden layers, and an output layer; the process of training the training samples based on the neural network to obtain the silicon wafer morphology prediction model includes: The process parameters are used as input parameters, and the silicon wafer thickness and morphology parameters are used as output parameters. Based on the input parameters and the output parameters, the number of neurons in the input layer, the hidden layer, and the output layer is determined to construct a BP neural network; Based on the input parameters and the output parameters, a silicon wafer morphology prediction model is obtained by training the training samples using a constructed BP neural network.

4. The method according to claim 3, characterized in that, The step of obtaining a silicon wafer morphology prediction model by training the training samples using a constructed BP neural network based on the input parameters and the output parameters includes: The input parameters are used as input variables of the input layer of the BP neural network. During the forward propagation process, the input variables undergo nonlinear transformation through the hidden layer to obtain the output variable value, and the output variable value is output by the output layer. Calculate the error between the output variable value and the theoretically set output parameters; If the error value is greater than the expected error, the weight and threshold of each neuron in the BP neural network are adjusted through the backpropagation algorithm until the error value is less than the expected error, thereby obtaining the silicon wafer morphology prediction model.

5. The method according to claim 2, characterized in that, The step of adjusting the current process parameters based on the predicted values ​​of silicon wafer thickness and morphology parameters and the set values ​​of silicon wafer thickness and morphology parameters conforming to the process standards, so that the predicted values ​​of silicon wafer thickness and morphology parameters meet the set values ​​of silicon wafer thickness and morphology parameters, includes: If the predicted value of the average thickness of the silicon wafer does not meet the set value of the average thickness of the silicon wafer, the initial thickness parameter in the current sensor thickness measurement parameters is adjusted. If the predicted value of the total thickness deviation of the silicon wafer does not meet the set value of the total thickness deviation of the silicon wafer, the grinding disc tilt angle parameter in the current process parameters will be adjusted.

6. The method according to claim 4, characterized in that, The method further includes: after the current round of training of the BP neural network ends, if the error value is greater than the expected error, then update the training steps of the BP neural network and start a new round of training.

7. A device for predicting and compensating the morphology of silicon wafer grinding, characterized in that, The device includes: The sample acquisition module is used to obtain training samples based on the process parameters affecting the grinding quality obtained during the historical grinding process of the silicon wafer and the thickness and morphology parameters of the silicon wafer measured after grinding. The process parameters include: grinding process parameters, grinding disc tilt angle parameters, grinding temperature parameters, grinding water flow rate parameters, and sensor thickness measurement parameters. The training module is used to train the training samples based on a neural network to obtain a silicon wafer morphology prediction model; The prediction module is used to input the current process parameters obtained during the current grinding process of the silicon wafer into the silicon wafer morphology prediction model to obtain the predicted values ​​of silicon wafer thickness and morphology parameters. The adjustment module is used to adjust the current process parameters based on the predicted values ​​of the silicon wafer thickness and morphology parameters and the set values ​​of the silicon wafer thickness and morphology parameters that meet the process standards, so that the predicted values ​​of the silicon wafer thickness and morphology parameters meet the set values ​​of the silicon wafer thickness and morphology parameters.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 6.

Images