Silicon carbide polishing solution precise regulation and control system based on multi-parameter feedback

Abstract

The invention discloses a silicon carbide polishing solution precise regulation and control system based on multi-parameter feedback, and particularly relates to the technical field of silicon carbide machining. Comprising a silicon carbide polishing solution central control module, a silicon carbide polishing solution parameter detection module, a silicon carbide polishing solution primary regulation and control module, a silicon carbide polishing solution secondary regulation and control module and a silicon carbide polishing solution regulation and control optimization module. The silicon carbide polishing solution central control module is used for receiving acquired data of the parameter detection module, generating a primary regulation and control instruction and a secondary regulation and control instruction, and driving the optimization module to perform parameter optimization; according to the method, a two-stage progressive regulation and control mechanism combining primary regulation and secondary regulation and control is adopted, and the primary regulation and control are mainly used for regulating basic characteristic parameters of the polishing solution so as to compensate basic characteristic deviation; and the secondary regulation and control is combined with polishing effect feedback to regulate the process adaptation parameters so as to correct the process adaptation deviation, so that the regulation and control precision and efficiency are improved.

Description

Technical Field

[0001] This invention relates to the field of silicon carbide processing technology, specifically to a precision control system for silicon carbide polishing slurry based on multi-parameter feedback. Background Technology

[0002] Silicon carbide (SiC), as a third-generation semiconductor material, has excellent properties such as high thermal conductivity, high breakdown field strength, and high electron mobility, and is widely used in high-end fields such as new energy vehicles, aerospace, and 5G communications. The polishing quality of silicon carbide substrates directly affects the fabrication performance of subsequent devices, and the characteristic parameters of polishing slurry are the core factors that determine the polishing quality.

[0003] Traditional silicon carbide polishing slurry control often employs single-parameter manual adjustment or static preset modes. For example, if low polishing efficiency is found, the abrasive content in the polishing slurry is increased; only the pH value of the polishing slurry is considered, and pH value is changed by adding acid-base adjusters to adapt to the polishing requirements of silicon carbide materials, assuming that a suitable pH value can ensure the stability of the polishing process; the formula of the polishing slurry, including the proportion of various components, is set in advance based on experience, without adjustment according to the actual processing conditions, thereby achieving control of the silicon carbide polishing slurry.

[0004] In practical applications, traditional silicon carbide polishing slurry control still suffers from the following shortcomings: First, the control parameters are singular, focusing only on a few indicators such as solid content or pH value, neglecting the impact of key parameters such as particle size distribution and zeta potential on the polishing effect; second, there is a lack of real-time feedback mechanism, resulting in a delayed control response, and the slurry characteristics cannot be compensated for in a timely manner when they change during the processing; third, a hierarchical control logic has not been formed, and deviations in basic characteristics and process adaptations interfere with each other, leading to low parameter matching accuracy; fourth, the control parameters mostly rely on empirical settings, lacking scientific optimization methods, making it difficult to achieve optimal polishing results. Therefore, developing a precision control system based on multi-parameter feedback is of great significance for improving the polishing quality of silicon carbide. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a precision control system for silicon carbide polishing slurry based on multi-parameter feedback, in order to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a precision control system for silicon carbide polishing slurry based on multi-parameter feedback, comprising:

[0007] The central control module for silicon carbide polishing slurry is used to receive data collected by the parameter detection module, generate primary and secondary control commands, drive the optimization module to optimize parameters, store data generated during the control of silicon carbide polishing slurry, and build a central control database for silicon carbide polishing slurry.

[0008] Silicon carbide polishing slurry parameter detection module: used to collect basic characteristic parameters, process state parameters and polishing effect parameters of polishing slurry in real time, construct silicon carbide polishing slurry detection dataset, and transmit it to the silicon carbide polishing slurry primary control module and silicon carbide polishing slurry secondary control module;

[0009] Primary control module for silicon carbide polishing slurry: It judges the allowable range of basic characteristic parameters of the polishing slurry in the test data of silicon carbide polishing slurry, calculates the deviation of each parameter from the standard value based on the judgment of abnormal results, adjusts according to the deviation, evaluates the adjustment results, and transmits the qualified adjustment evaluation results to the secondary control module of silicon carbide polishing slurry.

[0010] Secondary control module for silicon carbide polishing slurry: Based on the qualified results of the primary control evaluation, the module judges the allowable range of polishing effect parameters in the silicon carbide polishing slurry test dataset. It calculates the deviation of each parameter from the standard value based on the abnormal results, controls the process state parameters based on the deviation, evaluates the control results, and provides feedback based on the abnormal evaluation results to obtain the qualified results of secondary control of silicon carbide polishing slurry.

[0011] Silicon carbide polishing slurry regulation and optimization module: Based on the parameters generated during the regulation of silicon carbide polishing slurry, an optimization algorithm is used to generate an optimization scheme for the regulation parameters, which is then fed back to the central control module.

[0012] The technical effects and advantages of this invention are as follows:

[0013] 1. This invention can collect and process the basic characteristic parameters, process state parameters and polishing effect parameters of the polishing slurry in real time, accurately understand the state of the polishing slurry and the polishing effect, thereby achieving more targeted and comprehensive control, timely adjustment of polishing slurry parameters to adapt to different processing needs, and improve polishing quality and efficiency.

[0014] 2. This invention adopts a two-stage progressive control mechanism that combines primary control and secondary control. Primary control mainly adjusts the basic characteristic parameters of the polishing fluid to compensate for deviations in basic characteristics. Secondary control, in conjunction with polishing effect feedback, adjusts the process adaptation parameters to correct process adaptation deviations, thereby improving the accuracy and efficiency of control.

[0015] 3. This invention, through an optimization module, uses a hybrid algorithm combining particle swarm optimization and BP neural network based on historical control data and real-time feedback parameters to iteratively optimize the control parameters. This not only improves the accuracy and stability of control, but also reduces the need for manual intervention, lowers the difficulty and cost of operation, and improves the flexibility and versatility of the system. Attached Figure Description

[0016] Figure 1This is a schematic diagram of the overall process of the present invention.

[0017] Figure 2 This is a schematic diagram of the secondary control module for silicon carbide polishing slurry of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figure 1 As shown, the present invention provides a precision control system for silicon carbide polishing slurry based on multi-parameter feedback, including a silicon carbide polishing slurry central control module, a silicon carbide polishing slurry parameter detection module, a silicon carbide polishing slurry primary control module, a silicon carbide polishing slurry secondary control module, and a silicon carbide polishing slurry control optimization module.

[0020] The central control module of the silicon carbide polishing slurry is connected to all other modules. The silicon carbide polishing slurry parameter detection module is connected to the primary regulation module of the silicon carbide polishing slurry. The secondary regulation module of the silicon carbide polishing slurry is connected to both the primary regulation module and the regulation and optimization module of the silicon carbide polishing slurry. The regulation and optimization module of the silicon carbide polishing slurry is connected to the central control module of the silicon carbide polishing slurry.

[0021] The central control module for silicon carbide polishing slurry is used to receive data collected by the parameter detection module, generate primary and secondary control commands, drive the optimization module to optimize parameters, store data generated during the control of silicon carbide polishing slurry, and build a central control database for silicon carbide polishing slurry.

[0022] This embodiment requires specific explanation of the data generated during the control of silicon carbide polishing slurry, including but not limited to the collected raw data, processed data, control commands, and historical control records.

[0023] Silicon carbide polishing slurry parameter detection module: Used to collect basic characteristic parameters, process state parameters, and polishing effect parameters of the polishing slurry in real time, construct a silicon carbide polishing slurry detection dataset, and transmit it to the silicon carbide polishing slurry primary control module and the silicon carbide polishing slurry secondary control module, including:

[0024] A1: The basic characteristic parameters of the polishing slurry include particle size distribution, solid content, pH value, viscosity and zeta potential; the process state parameters of the polishing slurry include flow rate, temperature and spray pressure; the polishing effect parameters include polished surface roughness, silicon carbide substrate material removal rate and polished surface shape error.

[0025] This embodiment requires specific explanation regarding the integration of a particle size analyzer, solid content analyzer, pH meter, viscometer, zeta potential meter, flow sensor, temperature sensor, pressure sensor, speed sensor, optical profilometer, and laser interferometer into the parameter detection module to ensure detection accuracy. The laser interferometer can calibrate the detection results of the optical profilometer in real time. The particle size analyzer is used to detect the particle size distribution of abrasives in the polishing slurry; the solid content analyzer is used to detect the solid content of the polishing slurry; the pH meter is used to detect the pH value of the polishing slurry; the viscometer is used to detect the viscosity of the polishing slurry; the zeta potential meter is used to detect the zeta potential of the polishing slurry; the flow sensor is used to detect the supply flow rate of the polishing slurry; the temperature sensor is used to detect the real-time temperature of the polishing slurry; the pressure sensor is used to detect the spray pressure of the polishing slurry; and the optical profilometer is used to detect the surface roughness and surface shape error of the silicon carbide substrate after polishing. The silicon carbide substrate material removal rate Rr is obtained by comparing the difference between the thickness of the silicon carbide substrate before polishing and the thickness of the silicon carbide substrate after polishing with the polishing time point.

[0026] A2: Min-Max normalization was applied to the collected particle size distribution and corresponding standard values, solid content and corresponding standard values, pH value and corresponding standard values, viscosity and corresponding standard values, zeta potential and corresponding standard values, flow rate and corresponding standard values, temperature and corresponding standard values, injection pressure and corresponding standard values, circulation rate and corresponding standard values, surface roughness and corresponding standard values, removal rate and corresponding standard values, and surface shape error and corresponding standard values ​​to obtain the normalized parameter values ​​X and the corresponding standard values ​​X' of each parameter. th The threshold range for each parameter in the Min-Max normalization method is (X min ,X i,max The upper and lower limits of the allowable range for each parameter are (X). std,min ,X std,max ) and belong to (X) min ,X max The particle size distribution uses D50 (median particle size) as the core characteristic value. D50 is the particle size value that accounts for 50% of the total particle size distribution (e.g., D50 = 2 μm). Taking the viscosity of the polishing slurry as an example, the normalized viscosity GH and the corresponding standard value GH are obtained. th , , Viscosity sampling value is X, standard value is X std The viscosity threshold is (40, 60) MPa·s, the collected value is X = 55 MPa·s, the normalized GH is 0.75, and the standard value is X. std =50 MPa·s, normalized GH th=0.5, deviation ΔGH=0.75-0.5=0.25, indicating a range that deviates from the standard value by 25%; finally, the normalized parameters are used to construct a silicon carbide polishing slurry detection dataset;

[0027] In this embodiment, it should be specifically noted that the standard values ​​of each parameter are based on historical data. The initial standard values ​​are determined through orthogonal experiments, and then iteratively updated through an optimization module with the goal of achieving the best polishing effect. The allowable range of each parameter is centered on the standard value, and the offset is set according to the parameter adjustment precision and effect sensitivity, using the formula: X std,min =X std -ΔX,X std,min =X std +ΔX,X std ΔX and ΔX represent the standard value and offset of the parameters, respectively. This invention employs precise control, with ΔX ranging from 2% to 5% of the standard value. The threshold range of each parameter represents the system's "safety red line," relying on the hardware limits of the core equipment, material tolerance characteristics, and the critical point of process failure. Once calibrated, adjustments are only made during equipment replacement or process upgrades. The fundamental characteristic parameters are calibrated based on the chemical / physical tolerance limits of the material, such as zeta potential: experimentally tested critical potential for abrasive agglomeration (agglomeration intensifies at -30mV) and critical potential for excessive dispersion (-50mV). The abrasive settling temperature is calibrated to [-50, -30] mV. The process parameters are calibrated according to the operating limits of the equipment hardware. For example, the flow rate is calibrated to the minimum stable output (1.0 L / min, below which the pump runs dry) and maximum safe output (3.0 L / min, above which the motor is overloaded) of the polishing slurry delivery pump, calibrated to [1.0, 3.0] L / min. The polishing effect parameters are calibrated according to the workpiece scrap / process failure threshold, for example, the surface roughness is calibrated to the following: when R_a > 2.0 nm, the workpiece cannot meet the subsequent photolithography requirements, and when R_a < 0.5 nm, the polishing efficiency drops sharply, calibrated to [0.5, 2.0] nm.

[0028] This embodiment needs to specifically explain that the standard value anchors the optimal polishing effect; the allowable fluctuation range provides process tolerance space; the threshold ensures equipment and process safety. Without the threshold, the parameters may get out of control to the point of damaging the equipment or scrapping the workpiece.

[0029] The primary control module for silicon carbide polishing slurry: This module determines the allowable range of basic characteristic parameters of the polishing slurry in the test data set, calculates the deviation of each parameter from the standard value based on the anomaly detection results, adjusts the slurry according to the deviation, evaluates the adjustment results, and transmits the qualified adjustment results to the secondary control module for silicon carbide polishing slurry, including:

[0030] B1: Based on the raw value of the i-th parameter corresponding to the basic characteristics of the silicon carbide polishing slurry in the detection data, determine whether it belongs to the allowable range (X) of the i-th parameter.i,std,min ,X i,std,max If it belongs to the category, no adjustment is needed; otherwise, it indicates an abnormal judgment and primary PID control should be performed.

[0031] B2: First, based on the judgment of abnormal results, calculate the deviation Δx between the i-th parameter and the corresponding standard value. i Δx i =x i -x i，std x i and x i，std Let be the normalized sampled value and the standard value of the i-th parameter, respectively. If only one parameter has a deviation, then primary PID control is directly applied to this parameter. If at least two parameters have deviations, then the deviation Δx is used as the basis for further control. i The parameters are sorted by magnitude for primary PID control (e.g., a pH normalization deviation of +0.4 (40% of the range) is much larger than a solids deviation of -0.15 (15% of the range), indicating that pH should be prioritized for control); then, primary PID control of the corresponding component is applied to the i-th parameter to obtain the adjustment amount U of the i-th parameter. i (t), t is the parameter sampling period, K i,p K i,j and K i,d Let and be the proportional coefficient, integral coefficient, and differential coefficient of the i-th parameter, respectively;

[0032] In this embodiment, it should be specifically noted that the proportional coefficient, integral coefficient, and derivative coefficient of each parameter can be obtained from historical data according to a predetermined K. i,p Then adjust K i,j Finally, K is paired with i,d The order is determined, for example, by using the experimental controlled variable method, first calibrating the proportionality coefficient K. p Let the integral coefficient K j =0 and differential coefficient K d =0, establishing a linear relationship between the deviations of each parameter and the corresponding adjustment amount, until the adjustment amount is found so that the parameter value falls within the allowable range. The ratio of the adjustment amount to the parameter deviation is the proportional coefficient K. p (For example, if the deviation is 0.02 and the adjustment amount is 0.1, then the proportional coefficient K) i,p (0.1 / 0.02=0.5), then based on the obtained proportionality coefficient K p Let the differential coefficient K d =0, similarly creating deviations in each parameter, and setting the integral coefficient K. j Given the number of sampling periods m, the corresponding adjustment ratio K is obtained. p and integral coefficient K j Find the differential coefficient K dAfter adjustment, the integral coefficient K that falls within the allowable range of the parameter was found. j For example, basic characteristic parameters such as particle size distribution, solid content, pH value, viscosity, and Ka of zeta potential. p / K j / K d The values ​​can be 0.6 / 0.3 / 0.2, 0.8 / 0.4 / 0.1, 1.2 / 0.2 / 0.5, 0.7 / 0.3 / 0.3, and 1.0 / 0.1 / 0.4, respectively.

[0033] This embodiment requires specific explanation of the primary control equipment, which includes an abrasive supply component, a pH adjustment component, a viscosity adjustment component, and a dispersion stabilization component. The abrasive supply component adjusts the particle size distribution and solid content of the polishing slurry. The pH adjustment component adds acid or alkali solution to the polishing slurry via a precision titration pump, and works with a stirrer to ensure uniform mixing, thereby achieving precise pH adjustment. The viscosity adjustment component replenishes solvent or thickener via a metering pump based on viscosity deviation, quickly correcting viscosity parameters. The dispersion stabilization component uses an ultrasonic disperser to break up abrasive agglomerates, and combines this with a magnetic field stabilizer to maintain the abrasive dispersion state, adjusting the zeta potential to a stable range.

[0034] B3: Adjustment amount U based on the i-th parameter i (t), to evaluate the control results, first calculate the relative deviation between the actual value after control and the standard value, and obtain the deviation coefficient C of the i-th parameter. i , X i and X i,th Let X be the actual collected value and the standard value of the i-th parameter, respectively. i,max -X i,min The threshold range is the difference between the upper and lower limits of the i-th parameter; then the i-th parameter X is detected m times consecutively (e.g., 3 times, with a 1-second interval between each time). k Calculate the stability coefficient S of the i-th parameter. i , Next, the overall fit FC of the primary PID control is calculated. Where i = 1 to 5 represent particle size distribution, solid content, pH value, viscosity, and zeta potential, respectively. Finally, if FC is of excellent grade (e.g., [0.9, 1.0]), it means that the control evaluation result is qualified and meets the secondary control requirements of the polishing fluid. If FC is of good grade (e.g., [0.7, 0.9]), it means that the control evaluation result requires further adjustment of the parameters by 1 / 3. If FC is of average grade (e.g., [0.5, 0.7]), it means that the control evaluation result is unqualified, indicating that there is an abnormal result where the parameter deviation is greater than the threshold range. Repeat the control steps for the parameter with abnormal target deviation. If FC is of poor grade (e.g., [0, 0.5]), it means that the control evaluation result is unqualified. Stop the primary PID control and conduct root cause investigation, including investigation of the primary control equipment and investigation of basic characteristic parameters.

[0035] Please see Figure 2 As shown, the secondary control module for silicon carbide polishing slurry: Based on the qualified results of the primary control evaluation, it judges the allowable range of polishing effect parameters in the silicon carbide polishing slurry detection dataset, calculates the deviation of each parameter from the standard value based on the judgment of abnormal results, adjusts the process state parameters according to the deviation, evaluates the control results, and provides feedback based on the abnormal evaluation results to obtain the qualified results of the secondary control of silicon carbide polishing slurry, including:

[0036] C1: Determine whether the original values ​​of all parameters corresponding to the polishing effect in the silicon carbide polishing slurry detection dataset belong to the allowable range (X) of each parameter. std,min ,X std,max If all of them are true, no adjustment is needed; otherwise, calculate the deviation of the removal rate and surface error parameters in the polishing effect from the corresponding standard values, and prioritize the adjustment of the polishing effect parameters according to the magnitude of the deviation.

[0037] C2: When an anomaly is determined based on the removal rate, calculate the deviation ΔRr between the removal rate Rr and the corresponding standard value, where ΔRr = Rr - Rr th Rr and Rr th These are the normalized collected and standard values ​​of the removal rate, respectively; then, using historical data, the flow-removal rate relationship model was experimentally calibrated: Q th =Q+K Q ×|ΔRr|×sign(-ΔRr), K Q Here, Q is the flow regulation coefficient, and sign(-ΔRr) is the coincidence function, ensuring that the flow regulation direction matches the deviation direction. thThe normalized flow rate is collected as the sampled value and the standard value, respectively. Next, the standard flow rate output by the model is multiplied by the corresponding flow rate threshold range and then converted into a 4-20mA current signal (for example, through the inverse conversion of normalized value to physical quantity (e.g., 0-1 corresponds to 0-10L / min), the physical flow rate adjustment ΔQ = 0.25 × 10 = 2.5L / min, 4mA corresponds to 1L / min, and 20mA corresponds to 3L / min), which is then transmitted to the electric proportional valve. When ΔRr < 0, the current signal is increased, the valve opening increases, and the flow rate increases; when ΔRr > 0, the current signal is decreased, the valve opening decreases, and the flow rate decreases. Finally, the removal rate after regulation is detected and evaluated. If the removal rate after regulation falls within the corresponding allowable range, regulation is stopped, indicating that the regulation is qualified, and the current flow rate is maintained; otherwise, this step continues.

[0038] In this embodiment, it should be specifically noted that the 4-20mA current signal is an international standard signal in the field of industrial automation, which is formulated by the International Electrotechnical Commission (IEC) and is widely used in process control systems. Based on the removal rate deviation (ΔRr), the opening of the electric proportional valve is adjusted by changing the magnitude of the 4-20mA current signal, thereby controlling the polishing fluid flow rate and ultimately keeping the removal rate deviation within the allowable range.

[0039] C3: When the surface shape error in the polishing effect parameters is abnormal, collect the surface shape error (normalized surface shape error Wt) at h×h points (e.g., 200×200). Multiply the surface shape error Wt by the corresponding threshold range to obtain the physical quantity Wt of the surface shape error, generate a surface shape error heatmap, and divide the substrate into 3 regions according to the error value. The high error region is where Wt > the upper limit of the corresponding allowable range Wt. std,max The normal zone is the lower limit of the allowable zone, Wt. std,min ≤Wt≤Wt std,max The low error region is Wt≤Wt std,min Then, for errors in different regions, the injection pressure compensation P in the high-error zone and the low-error zone is calculated. 区 P 区 =P 基 +Ky×(Wt 区 -Wt th ), P 基 The reference pressure corresponds to the injection pressure value Wt in the normal zone. std,min (e.g., 0.3 MPa), Ky is the pressure compensation coefficient (calibrated experimentally, e.g., 0.2 MPa / μm), Wt 区 The average surface shape error of this region (e.g., Wt = 0.7 μm in high error areas), Wt th The standard value for surface shape error is used; then, based on the injection pressure compensation amount in the high error zone, a boost command (e.g., P) is sent to the corresponding pump set. gao=0.3+0.2×(0.7-0.4)=0.36MPa), based on the injection pressure compensation amount in the low error zone, a pressure reduction command is sent to the corresponding pump group. The injection pressure is increased or decreased by increasing / decreasing the motor frequency. Based on the normal zone, the current injection pressure and motor frequency are saved. Finally, the shape error distribution after regulation is detected and evaluated. That is, the area ratio of the high error zone and the low error zone is less than or equal to the threshold (e.g., less than or equal to 5%), and the average Wt of the polished whole surface is within the allowable range, the injection pressure regulation is stopped, indicating that the regulation is qualified. Otherwise, the process is repeated.

[0040] In this embodiment, it should be specifically noted that, through experimental calibration, the removal rate was found to be strongly correlated with the flow rate, and the surface shape error was strongly correlated with the injection pressure, based on the Pearson correlation coefficient.

[0041] C4: Compare the temperature T in the process status parameters with the standard value T th By performing a difference comparison, the temperature deviation ΔT is obtained, and the temperature adjustment amount U is obtained through the PID model. T (t), t is the parameter sampling period, K T,p K T,j and K T,d U and represent the proportional coefficient (e.g., 2.0), integral coefficient (e.g., 0.5), and derivative coefficient (e.g., 1.0) of the temperature regulation, respectively; when the original value corresponding to temperature T is greater than the upper limit of the corresponding allowable range, U T (t) Outputting a negative value triggers the cooling coil solenoid valve to open. When the original value corresponding to temperature T is less than the lower limit of the corresponding allowable range, U T (t) Output a positive value to trigger the heating jacket to be powered on. When the temperature T is within the allowable range, the heating jacket and cooling coil are in standby mode to maintain a stable temperature. At the same time, the speed sensor signal of the polishing fluid circulation pump is received to detect the speed and the corresponding polishing fluid temperature in real time. When the absolute difference between the polishing fluid temperature and the upper limit of the allowable range is less than the corresponding threshold (e.g., 0.5℃), it indicates that the speed causes the polishing fluid temperature to deviate from the allowable range, triggering the early activation of the cooling coil feedforward compensation mechanism. Finally, the polishing fluid temperature is detected and evaluated. If the adjusted polishing fluid temperature is within the corresponding allowable range, the adjustment is stopped and the current temperature is maintained, indicating that the adjustment is qualified. Otherwise, this step is continued.

[0042] In this embodiment, it should be specifically noted that the rotational speed of the polishing slurry circulation pump will generate frictional heat, which in turn will disturb the temperature of the polishing slurry. When the ambient temperature changes (e.g., from 25°C to 30°C), the heat dissipation efficiency will decrease, which may lead to the potential risk that the polishing slurry temperature will exceed the optimal range (e.g., 26°C).

[0043] C5: Based on the polished surface roughness in the polishing effect parameters, determine whether the original value collected is within the allowable range. If it is, it indicates that the secondary control evaluation of the silicon carbide polishing slurry is good. Otherwise, repeat steps C1-C4.

[0044] In this embodiment, it is necessary to specifically explain that the process parameters of flow rate, temperature and jet pressure all affect the surface roughness in different ways. Therefore, the surface roughness of the polished surface is used as a comprehensive factor to evaluate the secondary control effect of the polishing fluid.

[0045] Silicon carbide polishing slurry regulation and optimization module: Based on the parameters generated during the regulation of the silicon carbide polishing slurry, an optimization algorithm is used to generate an optimized regulation parameter scheme, which is then fed back to the central control module. This includes:

[0046] D1: Using the basic characteristic parameters and process state parameters from historical control data as input samples, and the corresponding polishing effect parameters as output samples, a BP neural network model is trained using the mean squared error loss function L. When L is less than a corresponding threshold (e.g., L < 10), the model is considered to be in a state where L is less than the corresponding threshold. --4 This indicates that the BP neural network model training is complete. , 3 represents the three parameters for polishing effect, y J and y J,m These are the actual value and the model output value of the J-th parameter, respectively;

[0047] D2: Minimize the fitness function fit with the goal of optimizing the polishing effect parameters. R a Wt, R r and R r,th These represent the normalized polished surface roughness, surface shape error, removal rate, and corresponding standard values, with a1, a2, and a3 representing the corresponding weights, e.g., a1=0.4, a2=0.3, and a3=0.3. Then, a particle swarm optimization algorithm is used to randomly generate D particles (e.g., 50 particles), each representing a combination of input samples. For each particle, its parameters are input into a trained BP neural network to obtain the predicted polishing effect parameters, which are then substituted into the fitness function to calculate fit. Next, iterative processing is performed using particle velocity update functions and particle position update functions. After each iteration... The individual optimal pbest and global optimal gbest of the input sample are updated. Finally, when the change in the fitness function value fit corresponding to the global optimal gbest is less than the corresponding threshold in n1 consecutive iterations (e.g., from fit 0.3256 to 0.3255, the change is 0.0001), the iteration stops. At this time, the global optimal gbest is the optimal combination of basic characteristic parameters and process state parameters. Each parameter in the optimal combination is denormalized to restore it to the actual physical quantity, and the optimization scheme of the control parameters is obtained. At the same time, it is fed back to the central control module.

[0048] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other.

[0049] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-parameter feedback based precision control system for silicon carbide polishing solution, characterized in that: The application relates to a carbonized silicon polishing liquid central control module, a carbonized silicon polishing liquid parameter detection module, a carbonized silicon polishing liquid primary regulation module, a carbonized silicon polishing liquid secondary regulation module and a carbonized silicon polishing liquid regulation optimization module. The carbonized silicon polishing liquid central control module is used for receiving collected data of the parameter detection module, generating primary regulation instructions and secondary regulation instructions, simultaneously driving the optimization module to perform parameter optimization, and storing data generated in the carbonized silicon polishing liquid regulation process and constructing a carbonized silicon polishing liquid central control database. The carbonized silicon polishing liquid parameter detection module is used for collecting basic characteristic parameters, process state parameters and polishing effect parameters of the polishing liquid in real time, constructing a carbonized silicon polishing liquid detection data set, and transmitting the carbonized silicon polishing liquid detection data set to the carbonized silicon polishing liquid primary regulation module and the carbonized silicon polishing liquid secondary regulation module. The carbonized silicon polishing liquid primary regulation module is used for judging the basic characteristic parameters of the polishing liquid in the carbonized silicon polishing liquid detection data set, calculating the deviation of each parameter from a standard value according to an abnormal judgment result, regulating according to the deviation, evaluating the regulation result, and transmitting the regulation evaluation qualified result to the carbonized silicon polishing liquid secondary regulation module. The carbonized silicon polishing liquid secondary regulation module is used for judging the polishing effect parameters in the carbonized silicon polishing liquid detection data set based on the primary regulation evaluation qualified result, calculating the deviation of each parameter from a standard value according to an abnormal judgment result, regulating the process state parameters according to the deviation, evaluating the regulation result, feeding back the abnormal evaluation result, and obtaining the carbonized silicon polishing liquid secondary regulation qualified result. The carbonized silicon polishing liquid regulation optimization module is used for generating a regulation parameter optimization scheme through an optimization algorithm based on the parameters generated in the carbonized silicon polishing liquid regulation process and feeding back the regulation parameter optimization scheme to the central control module. 2.The multi-parameter feedback based silicon carbide polishing solution precision control system of claim 1, wherein: The allowed interval range judgment in the primary regulation module of the silicon carbide polishing liquid: based on the original value corresponding to the i th parameter of the polishing liquid basic characteristics in the silicon carbide polishing liquid detection data set, whether it belongs to the allowed interval (X i,std,min , i,std,max ) of the i th parameter is judged, if it belongs to, no adjustment is needed, otherwise, the primary PID regulation is carried out. 3.The multi-parameter feedback based silicon carbide polishing solution precision control system of claim 1, wherein: The silicon carbide polishing liquid primary regulation module is regulated according to the deviation: first, based on the judgment of abnormal results, the deviation Δx of the i-th parameter and the corresponding standard value is calculated i , Δx i = x i - x i，std , x i and x i，std are the normalized acquisition value and the standard value of the i-th parameter, if only one parameter has a deviation, the secondary parameter is directly regulated by the primary PID, if at least two parameters have a deviation, the primary PID is regulated according to the size of the deviation Δx i Sort and regulate; then the i-th parameter is regulated by the corresponding component of the primary PID, and the adjustment amount U i (t) of the i-th parameter is obtained, , t is the parameter sampling period, K i,p , K i,j and K i,d are the proportional coefficient, integral coefficient and differential coefficient of the i-th parameter respectively. 4.The multi-parameter feedback based silicon carbide polishing solution precision control system of claim 1, wherein: The evaluation of the regulation result in the primary regulation module of the silicon carbide polishing solution is based on the adjustment amount U of the i-th parameter i (t), the evaluation of the regulation result is to calculate the relative deviation of the actual value after regulation and the standard value, and obtain the deviation coefficient C of the i-th parameter i , , X i and X i,th are the actual collected value and the standard value of the i-th parameter respectively, X i,max -X i,min is the difference between the upper and lower limit values of the i-th parameter, that is, the threshold range Then the i-th parameter X is detected continuously m times k , the stability coefficient S of the i-th parameter is calculated i , ; then the comprehensive fitness FC of the primary PID control is calculated , i=1 to 5 are particle size distribution, solid content, pH value, viscosity and zeta potential respectively; finally, if FC belongs to the excellent grade, it means that the control evaluation result is qualified and meets the secondary control requirements of the polishing liquid, if FC belongs to the good grade, it means that the control evaluation result needs to be adjusted by 1 / 3 of the adjustment amount, if FC belongs to the general grade, it means that the control evaluation result is unqualified, which means that there is an abnormal result with a parameter deviation greater than the threshold range, and the target abnormal parameter is repeatedly controlled, if FC belongs to the poor grade, it means that the control evaluation result is unqualified, and the primary PID control is stopped, and the root cause is investigated, including primary control equipment investigation and basic characteristic parameter investigation.

5. The multi-parameter feedback based precision control system for silicon carbide polishing solution of claim 1, wherein: The range of the allowed interval in the secondary control module of the silicon carbide polishing liquid is judged based on the original value of all parameters corresponding to the polishing effect in the silicon carbide polishing liquid detection data set, whether it belongs to the allowed interval of each parameter (X std,min ,X std,max ), if it belongs to the allowed interval, no adjustment is needed, otherwise, the deviation of the removal rate and the surface shape error parameters in the polishing effect from the corresponding standard value is calculated, and the polishing effect parameters are prioritized according to the deviation size. 6.The multi-parameter feedback based precision control system of a polishing solution for silicon carbide according to claim 1, wherein: The secondary regulation module of the silicon carbide polishing solution comprises: when the removal rate is judged to be abnormal, calculating the deviation ΔRr of the removal rate Rr and the corresponding standard value, ΔRr=Rr-Rr th , Rr and Rr th are the normalized collection values and standard values of the removal rate, respectively; then through historical data, an experimental calibration flow-removal rate relationship model is determined: Q th =Q+K Q ×|ΔRr|×sign(-ΔRr), K Q is a flow regulation coefficient, sign(-ΔRr) is a sign function, Q and Q th are the normalized collection values and standard values of the flow, respectively; then the flow standard value output by the model is multiplied by the corresponding flow threshold range and then converted into a 4-20mA current signal, which is transmitted to an electric proportional valve; when ΔRr<0, the current signal is increased, the valve opening is increased, and the flow is increased; when ΔRr>0, the current signal is decreased, the valve opening is decreased, and the flow is decreased; finally, the regulated removal rate is detected and evaluated; if the regulated removal rate belongs to the corresponding allowable range, the regulation is stopped, which indicates that the regulation is qualified, the current flow is maintained, otherwise, the step is continued.

7. The multi-parameter feedback based precision control system for silicon carbide polishing solution of claim 1, wherein: The secondary regulation module of the silicon carbide polishing liquid further comprises: when the surface error in the polishing effect parameter is abnormal, collecting the surface error of h x h points, multiplying the surface error Wt by the corresponding threshold range to obtain the physical quantity Wt of the surface error, generating a surface error thermodynamic map, and dividing the substrate into three regions according to the error value, wherein the high error region is Wt> the upper limit value Wt of the corresponding allowable interval std,max , the normal region is Wt std,min ≤Wt≤Wt std,max , and the low error region is Wt≤Wt std,min ; then, according to the error of different regions, calculating the jet pressure compensation amount P 区 of the high error region and the low error region, P 区 =P 基 +Ky×(Wt 区 -Wt th ), P 基 is the reference pressure, Wt std,min is the jet pressure value corresponding to the normal region, Ky is the pressure compensation coefficient, Wt 区 is the average surface error of the region, and Wt th is the standard value of the surface error; then, sending a pressure increase instruction to the corresponding pump group based on the jet pressure compensation amount of the high error region, sending a pressure decrease instruction to the corresponding pump group based on the jet pressure compensation amount of the low error region, and saving the current jet pressure and motor frequency based on the normal region; finally, detecting and evaluating the distribution of the surface error after regulation, that is, the area ratio of the high error region and the low error region is less than or equal to a threshold value, and the average Wt of the polishing full surface belongs to the allowable interval, the jet pressure regulation is stopped, which indicates that the regulation is qualified, otherwise the step is repeated. 8.The multi-parameter feedback based silicon carbide polishing solution precision control system of claim 1, wherein: The secondary regulation module of the silicon carbide polishing solution further comprises: regulating the temperature T in the process state parameters to a standard value T th The difference is compared to obtain the temperature deviation ΔT, and the temperature adjustment amount U is obtained by the PID model T (t); when the original value corresponding to the temperature T is greater than the upper limit value of the corresponding allowable interval, U T (t) outputs a negative value to trigger the opening of the cooling coil solenoid valve, and when the original value corresponding to the temperature T is less than the lower limit value of the corresponding allowable interval, U T (t) outputs a positive value to trigger the heating jacket to be powered on, and when the temperature T belongs to the allowable interval, the heating jacket and the cooling coil are in standby state to maintain the temperature stable; at the same time, the speed sensor signal of the polishing solution circulating pump is received to detect the speed and the corresponding polishing solution temperature in real time, and when the absolute difference between the polishing solution temperature and the upper limit value of the allowable interval is less than the corresponding threshold value, it is indicated that the speed causes the polishing solution temperature to develop in the direction deviating from the allowable interval, and the feedforward compensation mechanism of the cooling coil is triggered to be opened in advance; finally, the polishing solution temperature is detected and evaluated, and if the regulated polishing solution temperature belongs to the corresponding allowable interval, the regulation is stopped to maintain the current temperature, which indicates that the regulation is qualified, otherwise the step is continued.

Images