A stainless steel pipe inner wall polishing process optimization method and system

Abstract

The application provides a stainless steel pipe inner wall polishing process optimization method and system, and belongs to the technical field of intelligent optimization. The method comprises the following steps: sampling endoscopic detection is performed on the stainless steel pipe to obtain an inner wall roughness array, and initial polishing parameters of the polishing process are configured; polishing prediction is performed according to the initial polishing parameters and the inner wall roughness array to obtain an initial polishing roughness array, and vibration prediction is performed to obtain an initial polishing vibration parameter array; vibration acceptability verification is performed on the initial polishing vibration parameter array and the inner wall roughness array to obtain vibration acceptability; polishing parameter optimization step configuration is performed according to the initial polishing roughness array and the vibration acceptability to obtain a polishing parameter optimization step array, and adjustment optimization iteration is performed. The application solves the problems that the existing stainless steel pipe inner wall polishing process relies on manual experience and repeatedly tries and errors, lacks the ability of prior prediction and vibration evaluation, and leads to poor polishing uniformity and long debugging period.

Description

Technical Field

[0001] This invention relates to the field of intelligent optimization technology, specifically to a method and system for optimizing the polishing process of the inner wall of stainless steel pipes. Background Technology

[0002] The polishing quality of the inner wall of stainless steel pipes directly affects the performance of piping systems in high-cleanliness fields such as food, pharmaceuticals, and semiconductors. Excessive inner wall roughness not only increases fluid transport resistance but also easily fosters the growth of microorganisms or residual media, leading to product contamination risks. Therefore, efficient and uniform polishing of the inner wall of stainless steel pipes is a crucial step in ensuring their safe application.

[0003] Currently, polishing the inner walls of stainless steel pipes mainly relies on abrasive flow machining, electrochemical polishing, or mechanical honing. These processes typically rely on operator experience to set parameters such as abrasive type, pressure, flow rate, and processing time. However, due to the complex factors of stainless steel pipe diameter variations, bend distribution, and uneven initial roughness, the same parameter combination often fails to ensure uniform polishing across the entire pipe. To verify the effect, current practices often involve sampling and endoscopic inspection and roughness measurement of the pipe after polishing. If local areas are found to be below target values, parameters must be adjusted based on experience, and reprocessing is repeated through trial and error. Furthermore, vibration issues during polishing are often overlooked. Severe fluid or mechanical vibrations not only exacerbate equipment wear but also lead to localized over- or under-polishing. Current technologies lack systematic methods for effectively predicting and controlling vibration before processing.

[0004] In summary, the existing stainless steel pipe inner wall polishing process relies on manual experience and lacks the ability to predict and dynamically optimize in advance, resulting in long debugging cycles, poor parameter applicability, difficulty in ensuring the uniformity of the entire pipe, and the inability to quantitatively assess the impact of vibration before processing. Summary of the Invention

[0005] This invention provides a method and system for optimizing the polishing process of the inner wall of stainless steel pipes, which solves the problems of existing stainless steel pipe inner wall polishing processes relying on repeated trial and error based on manual experience and lacking the ability to predict and assess vibrations in advance, resulting in poor polishing uniformity and long debugging cycles.

[0006] In view of the above problems, the present invention provides an optimized method and system for polishing the inner wall of stainless steel pipes.

[0007] In a first aspect, the present invention provides an optimized method for polishing the inner wall of a stainless steel pipe, comprising: Sampling endoscopic inspection was performed on the stainless steel pipe to obtain the inner wall roughness array, and the initial polishing parameters for the polishing process were configured. Based on the initial polishing parameters and the inner wall roughness array, polishing prediction is performed to obtain the initial polishing roughness array, and vibration prediction is performed to obtain the initial polishing vibration parameter array. The vibration acceptability of the initial polishing vibration parameter array and the inner wall roughness array is verified to obtain the vibration acceptability. Based on the initial polishing roughness array and vibration tolerance, the polishing parameter optimization step size configuration is performed to obtain the polishing parameter optimization step size array. The optimization is then adjusted and iterated to obtain the optimal polishing parameters.

[0008] Secondly, the present invention provides an optimized system for polishing the inner wall of stainless steel pipes, comprising: The data acquisition and initial configuration module is used to perform sampling endoscopic inspection on stainless steel pipes to obtain an inner wall roughness array and configure the initial polishing parameters for the polishing process. The prediction and simulation module is used to perform polishing prediction based on the initial polishing parameters and the inner wall roughness array to obtain the initial polishing roughness array, and to perform vibration prediction to obtain the initial polishing vibration parameter array. The vibration acceptability verification module is used to verify the vibration acceptability of the initial polishing vibration parameter array and the inner wall roughness array to obtain the vibration acceptability. The optimization iteration and output module is used to configure the polishing parameter optimization step size according to the initial polishing roughness array and vibration acceptance, obtain the polishing parameter optimization step size array, and perform adjustment and optimization iteration to obtain the optimal polishing parameters.

[0009] One or more technical solutions provided in this invention have at least the following technical effects or advantages: The technical solution of this invention first obtains the roughness array of the inner wall of the stainless steel tube through sampling endoscopic detection, and automatically generates initial polishing parameters from the preset polishing parameter space based on the deviation ratio between the mean roughness value and the target value. This quantitative detection data replaces manual experience setting, thereby improving the rationality of the parameter starting point from the source.

[0010] Furthermore, by utilizing a pre-trained polishing prediction model and a pre-built polishing vibration database, the polishing roughness distribution and vibration amplitude under the current parameters are simultaneously estimated, enabling a two-dimensional prediction of polishing effect and vibration state before processing, thus avoiding ineffective trial polishing.

[0011] Furthermore, the roughness array and vibration parameter array are normalized and the vibration acceptance degree is calculated. The parameter combinations with excessively high vibration risk are screened out by setting a preset acceptance threshold, providing a safe boundary for subsequent optimization.

[0012] Finally, the deviation of each detection point is calculated based on the allowable roughness threshold, and the preset adjustment step size is corrected by a double coefficient in conjunction with the vibration acceptance. A variable step size search array is generated, and the multiple sets of candidate parameters after the variable step size adjustment are iteratively optimized by the particle swarm algorithm. After convergence, the optimal polishing parameters that take into account both roughness compliance and vibration coordination are output.

[0013] In summary, the technical solution of this invention realizes fully automated decision-making of the stainless steel pipe inner wall polishing process parameters from initial configuration to final optimization, significantly shortens the debugging cycle, reduces trial polishing costs, and ensures the polishing uniformity and processing stability of the entire pipe inner wall. This solves the problems of existing stainless steel pipe inner wall polishing processes relying on repeated trial and error based on manual experience and lacking the ability to predict and assess vibrations in advance, resulting in poor polishing uniformity and long debugging cycles. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating an optimized polishing process for the inner wall of a stainless steel pipe provided by the present invention.

[0015] Figure 2 This is a schematic diagram illustrating the process of generating optimal polishing parameters in a method for optimizing the polishing process of the inner wall of a stainless steel pipe provided by the present invention.

[0016] Figure 3 This is a schematic diagram of the structure of an optimized polishing process system for the inner wall of stainless steel pipes provided by the present invention.

[0017] In the attached diagram, the labels representing each component are as follows: The module includes a data acquisition and initial configuration module 11, a prediction and simulation module 12, a vibration acceptability verification module 13, and an optimization iteration and output module 14. Detailed Implementation

[0018] This invention provides an optimized method and system for polishing the inner wall of stainless steel pipes, which solves the problems of existing stainless steel pipe inner wall polishing processes that rely on repeated trial and error based on manual experience, lack of pre-prediction and vibration assessment capabilities, resulting in poor polishing uniformity and long debugging cycles.

[0019] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0020] Example 1, as Figure 1 As shown, the present invention provides an optimized method for polishing the inner wall of a stainless steel pipe, the method comprising: S100: Perform sampling endoscopic inspection on stainless steel pipes to obtain an inner wall roughness array and configure the initial polishing parameters for the polishing process.

[0021] This step obtains roughness distribution data of the inner wall of the stainless steel tube through sampling endoscopic inspection, and automatically configures the initial polishing parameters based on the inspection results and preset rules, providing an input basis for subsequent polishing prediction and parameter optimization.

[0022] Step S100 in the method provided by the present invention includes: Randomly sample stainless steel pipes from the same batch, and select test points in multiple preset directions within the sampled stainless steel pipes. Use industrial endoscopic equipment to perform internal roughness testing to obtain an internal roughness array. Obtain the polishing parameter space and select to generate initial polishing parameters, including the polishing speed.

[0023] In this step, random sampling is first performed on the stainless steel tubes to be processed in the same batch, according to a preset sampling ratio. For the selected sample tubes, inspection points are selected on their inner walls along multiple preset directions. The inner wall surface images of the above inspection points are acquired one by one using an industrial endoscope, the roughness values ​​are extracted, and the data are compiled to form an internal roughness array.

[0024] The sampling ratio is set according to the batch size and the required confidence level of the test. When the batch size is less than 100 samples, the sampling ratio is no less than 10% to ensure the representativeness of the samples in small-batch scenarios; when the batch size is greater than or equal to 100 samples, the sampling ratio is no less than 5% to maintain statistical validity while taking into account the testing cost.

[0025] The preset directions are evenly distributed along the circumference, with no fewer than four sections, to capture potential roughness differences in all directions around the circumference. Simultaneously, one or more cross-sections can be selected along the axial direction based on the pipe length; the number of cross-sections is set according to the pipe length and the required detection accuracy. For long straight pipes, the number of axial cross-sections is no fewer than two to cover roughness variations along the pipe's length. For simplicity, this embodiment uses only a single cross-section as an example; in actual implementation, the number of cross-sections can be increased as needed to more comprehensively characterize the roughness distribution of the entire pipe's inner wall.

[0026] For example, a batch contains 60 304 stainless steel pipes. Since the quantity is less than 100, the preset ratio is 10%, meaning 60 × 10% = 6 pipes are selected as sample pipes. For one straight pipe with an inner diameter of 50mm and a length of 1.5m, a cross-section is taken 800mm from the pipe opening. Four inspection points are taken along four preset directions: 0°, 90°, 180°, and 270°, for a total of four inspection points. After inspection with an industrial endoscope, the internal roughness array of the pipe is obtained as [0.9, 1.0, 1.3, 1.0], in μm.

[0027] It should be noted that the sampling method of this single cross section is only used to demonstrate the process flow. In actual implementation, multiple cross sections can be selected in the axial direction to fully cover the roughness distribution along the entire length of the pipe, thereby more accurately supporting the evaluation of the polishing uniformity of the entire inner wall of the pipe.

[0028] Furthermore, a pre-defined polishing parameter space is obtained. The pre-defined polishing parameter space refers to a multi-dimensional set of feasible parameter values ​​established in advance through offline experimental design or historical data mining before the formal execution of the polishing parameter optimization process. It defines the adjustable range and discrete value intervals for each polishing process parameter, providing a search boundary for subsequent initial parameter selection and iterative optimization.

[0029] The polishing parameter space is constructed as follows: for specific pipe materials and pipe diameter specifications, orthogonal experimental tables or uniform experimental tables are designed, polishing tests are carried out under different parameter combinations, the parameter range that can achieve qualified polishing effect and stable equipment operation is recorded, and abnormal boundary values ​​that cause over-polishing, under-polishing, excessive vibration or equipment alarm are eliminated, and the feasible upper and lower limits and step size of each parameter are formed.

[0030] For example, orthogonal experiments were conducted on an abrasive flow polishing machine for straight pipes made of 304 stainless steel with an inner diameter of 40-60 mm. The experiments revealed that grinding efficiency was too low below 1000 rpm, failing to meet the required speed within a reasonable time; above 1900 rpm, the pipe wall temperature rose drastically, and the vibration amplitude exceeded the equipment's safety threshold. Based on this, the feasible range for polishing speed of this pipe specification was set to 1000-1900 rpm, with a step size of 100 rpm. Similarly, the feasible range for abrasive concentration was set to 20%-50%, with a step size of 5%; the feasible range for fluid pressure was set to 0.3-0.7 MPa, with a step size of 0.05 MPa; and the feasible range for processing time was set to 10-30 minutes, with a step size of 2 minutes. These ranges and step sizes together constitute the preset polishing parameter space for this specific working condition.

[0031] Finally, based on the polishing parameter space, combined with the pipe material grade, pipe diameter, and statistical values ​​of the collected internal roughness array of the sample tube, a set of initial polishing parameters is automatically selected and generated according to a certain initialization strategy, such as the center point method, random sampling method, or rule-based matching method, including the key polishing speed parameter.

[0032] Preferably, the initial polishing parameters can be automatically generated using a rule-based matching method: First, extract the sample tube material grade, tube diameter, and the mean and maximum values ​​of the internal roughness array, and then retrieve the corresponding preset polishing parameter space according to the tube material and tube diameter.

[0033] Next, calculate the deviation ratio K = mean roughness / target roughness, proportionally map the interval between the minimum value and maximum value of each parameter in the polishing parameter space, and calculate the specific value of each parameter through linear interpolation P = minimum value + λ × (maximum value - minimum value). When K>3, it is mapped to the high value segment where λ=0.85; when 1.5<K≤3, it is mapped to the medium value segment where λ=0.55; when K≤1.5, it is mapped to the low value segment where λ=0.25. Finally, a set of initial polishing parameters including the polishing rotation speed is output.

[0034] For example, the inner diameter of the 304 stainless steel straight sample tube is 50mm, the mean value of the inner roughness array is 1.05μm, and the target roughness is 0.2μm. The calculated deviation ratio K=1.05 / 0.2=5.25>3, so λ=0.85 is selected. The corresponding polishing parameter space is retrieved: polishing rotation speed 1000 to 1900rpm, fluid pressure 0.3 to 0.7MPa, abrasive concentration 20% to 50%, processing time 10 to 30 minutes. Calculated according to λ=0.85: rotation speed=1000+0.85×(1900-1000)=1765rpm, pressure=0.3+0.85×(0.7-0.3)=0.64MPa, abrasive concentration=20%+0.85×(50%-20%)=45.5%, processing time=10+0.85×(30-10)=27 minutes. The final output initial polishing parameters are: rotation speed 1765rpm, pressure 0.64MPa, abrasive concentration 45.5%, processing time 27 minutes.

[0035] It should be noted that the above numerical values are only used for illustrative description and do not constitute a limitation on the present invention.

[0036] In summary, this step replaces manual empirical judgment with quantitative detection data, so that the setting of initial polishing parameters accurately corresponds to the actual roughness level of the sample tube, reduces blind trial and error from the source, effectively shortens the process debugging cycle and improves the polishing uniformity of the inner wall of the entire tube; when multiple detection sections are arranged in the axial direction, it can also support the evaluation of polishing uniformity in the length direction of the entire tube at the same time.

[0037] S200: Perform polishing prediction according to the initial polishing parameters and the inner wall roughness array to obtain an initial polishing roughness array, and perform vibration prediction to obtain an initial polishing vibration parameter array.

[0038] This step uses a pre-trained polishing prediction model and a pre-constructed polishing vibration database to perform polishing effect prediction and equipment vibration prediction on the current initial polishing parameters and inner wall roughness array respectively, and outputs an initial polishing roughness array and an initial polishing vibration parameter array.

[0039] In this step, polishing prediction is performed based on the initial polishing parameters and the inner wall roughness array to obtain the initial polishing roughness array, including: Based on the stainless steel pipe inner wall polishing test records, the sample polishing parameter set and the sample original inner wall roughness set were collected, as well as the inner wall roughness after polishing according to different sample polishing parameters and sample original inner wall roughness, to obtain the sample polishing roughness set. A model training and testing mechanism is adopted to divide the sample polishing parameter set, the sample original inner wall roughness set, and the sample polishing roughness set to obtain training data and test data. A qualified polishing prediction model is obtained by training and testing. The initial polishing parameters and each inner wall roughness in the inner wall roughness array are input into the polishing prediction model, and the initial polishing roughness array is output.

[0040] Specifically, based on the stainless steel pipe inner wall polishing test records, three sets of corresponding sample data were collected: a sample polishing parameter set, which is the combination of parameters such as rotation speed, pressure, abrasive concentration, and processing time used in each polishing process; a sample original inner wall roughness set, which is the inner wall roughness value of each pipe before polishing; and a sample polishing roughness set measured after actual polishing according to the above parameters and original roughness, which is the final inner wall roughness value of each pipe after polishing.

[0041] For example, 100 valid samples were collected from historical test records. In one sample, the polishing parameters were: polishing speed 1500 rpm, pressure 0.5 MPa, abrasive concentration 35%, processing time 20 minutes, the original inner wall roughness of the sample was 0.9 μm, and the final inner wall roughness after polishing was 0.15 μm.

[0042] Secondly, a model training and testing mechanism is employed. The collected sets of sample polishing parameters, original sample inner wall roughness, and sample polished roughness are randomly divided into training and testing datasets according to a preset ratio. The training dataset is used to fit the model's internal mapping parameters, while the testing dataset is used to evaluate whether the model's prediction accuracy is satisfactory after training.

[0043] The preset ratio refers to the proportion of training data to the total sample size, which is selected based on the sample size: 8:2 for samples ≥ 500 (80% training set, 20% test set); 7:3 for samples between 100 and 500; and leave-one-out cross-validation for samples less than 100. For example, 100 samples can be divided in a 7:3 ratio, with 70 samples used as training data and 30 as test data.

[0044] Next, an appropriate regression model, such as a backpropagation neural network or support vector regression, is selected. The polishing parameters and original inner wall roughness in the training dataset are used as input features, and the corresponding polished roughness is used as the output label for iterative training. After training, the model is validated using a test dataset. When the root mean square error between the predicted roughness and the actual roughness is lower than a preset threshold, the polishing prediction model is deemed qualified, training is stopped, and the model is saved.

[0045] Before training, the input data needs to be normalized to map parameters with different dimensions, such as rotation speed, pressure, and roughness, to the [0,1] interval. The specific process is as follows: for each feature dimension, its minimum and maximum values ​​in all samples of the training set are calculated. Then, a linear transformation is performed sample by sample using the formula: normalized value = (current value - minimum value) / (maximum value - minimum value), so that all feature values ​​fall within the [0,1] range. During the test set and subsequent predictions, the same transformation is performed using the minimum and maximum values ​​saved in the training set to maintain scale consistency. The rotation speed and pressure ranges used for normalization are consistent with the polishing parameter space.

[0046] For example, the minimum value of the rotational speed feature in the training set is 1000 rpm, and the maximum value is 1900 rpm. For a sample with a rotational speed of 1500 rpm, the normalized value is (1500-1000) / (1900-1000) = 500 / 900 ≈ 0.556. The pressure range is 0.3~0.7 MPa. For a sample with a pressure of 0.5 MPa, the normalized value is (0.5-0.3) / (0.7-0.3) = 0.5. And so on, the normalized input is [0.556, 0.5, 0.5, 0.5, 0.875].

[0047] Preferably, a backpropagation neural network is selected as the regression model, with the normalized polishing parameters and the original inner wall roughness in the training dataset as input features, and the corresponding polished roughness as the label output, for iterative training; the preset threshold is set according to the tolerance range of the target roughness, usually taking 1 / 5 to 1 / 3 of the target roughness tolerance, to ensure that the prediction error will not significantly affect the accuracy of subsequent parameter optimization.

[0048] For example, if the target roughness is 0.2μm and the allowable fluctuation range of the process is ±0.05μm, then the preset threshold can be set to 40% of the tolerance range, i.e., preset threshold = 0.05 × 40% = 0.02μm. This ensures that the prediction deviation of the polishing prediction model is much smaller than the allowable fluctuation of the process, and avoids the optimization direction being wrong due to inaccurate prediction.

[0049] For example, a three-layer BP neural network is used, with 5 nodes in the input layer corresponding to rotation speed, pressure, concentration, time, and original roughness, 10 nodes in the hidden layer, and 1 node in the output layer corresponding to the predicted roughness. After 500 iterations, the root mean square error on 30 sets of test data is 0.015μm, which is lower than the preset threshold of 0.02μm, so it is judged to be qualified, and the polishing prediction model is saved.

[0050] Finally, the initial polishing parameters generated in the previous steps are paired one by one with the roughness value of each detection point in the inner wall roughness array, and then input into the trained polishing prediction model. The model outputs the predicted roughness after polishing for each detection point, and the values ​​are summarized in the original detection point order to form the initial polishing roughness array.

[0051] For example, the initial polishing parameters are: rotation speed 1765 rpm, pressure 0.64 MPa, concentration 45.5%, and time 27 minutes. The roughness array of the inner wall of the sample tube consists of four points, with the original roughness of each point being [0.9, 1.0, 1.3, 1.0], in μm. After normalization, the initial polishing parameters yield the input data [0.804, 0.733, 0.85, 0.85, 0.4], corresponding to 0.9 μm at the first point. This is fed into the polishing prediction model, which outputs a predicted roughness of 0.12 μm after polishing. This process is repeated for the four points to obtain the initial polishing roughness array [0.12, 0.14, 0.16, 0.13], in μm, resulting in four predicted values.

[0052] In this step, vibration prediction is performed to obtain an initial polishing vibration parameter array, including: Access the polishing vibration database; The initial polishing parameters are used to search the polishing vibration database to obtain an initial polishing vibration parameter array, wherein the initial polishing vibration parameter array includes vibration amplitudes in multiple preset directions.

[0053] Specifically, when performing vibration prediction, a pre-built polishing vibration database is first invoked. The polishing vibration database stores the mapping relationship between different combinations of polishing parameters and the multi-directional vibration amplitude of the equipment, providing data support for vibration prediction.

[0054] The steps for constructing the polishing vibration database include: Collect a set of sample polishing parameters and test the vibration amplitude of the polishing equipment in multiple preset directions under different sample polishing parameters to obtain a set of sample vibration amplitude arrays; A mapping relationship is constructed between the sample polishing parameter set and the sample vibration amplitude array set, and fuzzy mapping rules are configured to build a polishing vibration database. The fuzzy mapping rules include matching the closest sample polishing parameters to the input polishing parameters for mapping.

[0055] Specifically, firstly, a set of sample polishing parameters that are from the same source as the polishing prediction model is collected. Then, each set of sample parameters is actually run on the polishing equipment. Using accelerometers installed on the main shaft bearing seat and pipe clamp of the equipment, the vibration amplitudes in multiple preset directions are tested and recorded respectively, thus obtaining a set of sample vibration amplitude arrays corresponding to the sample polishing parameters.

[0056] The preset vibration direction is consistent with the preset direction of the endoscopic detection in the previous step, namely the four directions of circumferential 0°, 90°, 180° and 270°, so that the vibration data corresponds to the spatial orientation of the roughness detection point, which facilitates the dimension-by-dimensional correlation calculation of the two vectors in the subsequent vibration acceptance verification.

[0057] For example, 100 sets of parameters from the sample polishing parameter set are taken and run on the equipment one by one. Among them, the parameters of sample A are a rotation speed of 1700 rpm and a pressure of 0.6 MPa. After the operation is stable, the vibration amplitudes in the four preset directions of the circumference are measured to be 0.12 mm, 0.10 mm, 0.08 mm and 0.10 mm respectively, which are recorded as a mapping record: [1700,0.6]-[0.12,0.10,0.08,0.10].

[0058] Next, a mapping relationship is established between the sample polishing parameter set and the sample vibration amplitude array set, forming the basic table of the vibration database. Simultaneously, fuzzy mapping rules are configured to address situations where the input parameters do not perfectly match the samples in the database during retrieval.

[0059] The fuzzy mapping rule specifically involves calculating the normalized Euclidean distance between the input parameters and the parameters of each sample in the database, selecting the sample with the smallest Euclidean distance as the matching result, and outputting its corresponding vibration amplitude array. If multiple samples have equal Euclidean distances, the group with the larger vibration amplitude value is selected to ensure the safety of subsequent vibration verification.

[0060] For example, the normalized Euclidean distance between the initial parameters [1765, 0.64] and sample A [1700, 0.6] is 0.12, and the distance between the initial parameters and sample B [1800, 0.7] is 0.15. The rotational speed and pressure range used in the calculation of the normalized Euclidean distance are consistent with the polishing parameter space, i.e., rotational speed 1000~1900rpm, pressure 0.3~0.7MPa. Sample A with a smaller distance is selected for mapping, and the output vibration amplitude is [0.12, 0.10, 0.08, 0.10], in mm.

[0061] Finally, the mapping relationship table is integrated with the fuzzy mapping rules to form a complete polishing vibration database, which can be used for vibration prediction.

[0062] Finally, the initial polishing vibration parameter array is obtained by searching the polishing vibration database using the initial polishing parameters. For example, using the initial polishing parameters of 1765 rpm and 0.64 MPa from the previous example, the normalized Euclidean distance with each sample in the polishing vibration database is calculated. Sample A has the smallest Euclidean distance, indicating a successful match, and the initial polishing vibration parameter array [0.12 mm, 0.10 mm, 0.08 mm, 0.10 mm] is output.

[0063] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0064] In summary, this step can simultaneously predict the polishing quality distribution and equipment vibration status without actual test polishing, providing a quantitative basis for subsequent vibration acceptability verification and parameter optimization, significantly shortening the debugging cycle and reducing test polishing costs.

[0065] S300: The initial polishing vibration parameter array and the inner wall roughness array are subjected to vibration acceptability verification to obtain the vibration acceptability.

[0066] This step normalizes the inner wall roughness array and the initial polishing vibration parameter array, respectively, to construct a normalized roughness vector and a normalized vibration vector. The vibration acceptability of the two is calculated and compared with the preset acceptability threshold to verify the feasibility of the current initial polishing parameters at the vibration level.

[0067] Step S300 in the method provided by the present invention includes: The inner wall roughness array is normalized to obtain a normalized roughness array; The initial polishing vibration parameter array is normalized to obtain a normalized vibration parameter array; The vibration acceptability was calculated and verified by the normalized roughness array and the normalized vibration parameter array to obtain the vibration acceptability.

[0068] In this step, each roughness value in the inner wall roughness array is first linearly transformed according to the minimum and maximum roughness values ​​saved when training the polishing prediction model, and uniformly mapped to the [0,1] interval to obtain a normalized roughness array, so as to eliminate the interference of roughness dimension differences on subsequent vibration verification calculations.

[0069] For example, the roughness range saved during training is 0.5 to 1.5 μm. The four points of the inner wall roughness array are [0.9, 1.0, 1.3, 1.0], and after normalization: point 1 = (0.9-0.5) / (1.5-0.5) = 0.4, point 2 = (1.0-0.5) / 1.0 = 0.5, point 3 = (1.3-0.5) / 1.0 = 0.8, point 4 = (1.0-0.5) / 1.0 = 0.5, resulting in the normalized roughness array [0.4, 0.5, 0.8, 0.5].

[0070] Furthermore, the vibration amplitudes in each direction of the initial polishing vibration parameter array are linearly transformed according to the minimum and maximum values ​​of the vibration amplitudes in each direction stored in the polishing vibration database, and uniformly mapped to the [0,1] interval to obtain a normalized vibration parameter array, so as to perform vibration acceptance verification calculations on a unified scale with the normalized roughness array.

[0071] For example, the vibration amplitude range in the four circumferential directions in the vibration database is 0.03 to 0.30 mm. After normalizing the initial vibration parameter array [0.12, 0.10, 0.08, 0.10]: 0° direction = (0.12-0.03) / (0.30-0.03) = 0.333, 90° direction = (0.10-0.03) / 0.27 = 0.259, 180° direction = (0.08-0.03) / 0.27 = 0.185, 270° direction = (0.10-0.03) / 0.27 = 0.259, the normalized vibration parameter array [0.333, 0.259, 0.185, 0.259] is obtained.

[0072] Furthermore, the vibration acceptability of the normalized roughness array and the normalized vibration parameter array is calculated and verified to obtain the vibration acceptability, including: Based on the normalized roughness array and the normalized vibration parameter array, a normalized roughness vector and a normalized vibration vector are constructed. Calculate the similarity between the normalized roughness vector and the normalized vibration vector to obtain directional consistency; Calculate the vibration safety factor based on the normalized vibration vector; Multiplying the directional consistency by the vibration safety factor yields the vibration acceptability.

[0073] Specifically, based on the obtained normalized roughness array and normalized vibration parameter array, normalized roughness vector and normalized vibration vector are constructed respectively. The normalized roughness vector is composed of the normalized roughness values ​​of each detection point arranged in order, and its dimension is equal to the total number of detection points; the normalized vibration vector is composed of the normalized vibration amplitudes of each preset direction arranged in the same direction, and the two vectors have the same dimension and each dimension corresponds one-to-one.

[0074] For example, the normalized roughness array has four values, forming a 4-dimensional normalized roughness vector [0.4, 0.5, 0.8, 0.5]. The normalized vibration parameter array is also 4-dimensional, forming a normalized vibration vector [0.333, 0.259, 0.185, 0.259]. The two vectors have the same dimensions and each dimension corresponds one-to-one, that is, the roughness in the 0° direction corresponds to the vibration in the 0° direction, and so on.

[0075] Next, the cosine similarity between the normalized roughness vector and the normalized vibration vector is calculated as a measure of directional consistency, used to assess the degree of shape matching between the roughness distribution pattern and the vibration distribution pattern. The cosine similarity formula is a well-known formula and will not be elaborated here.

[0076] For example, the cosine similarity is used to calculate the similarity between the normalized roughness vector and the normalized vibration vector, that is, to calculate the cosine value of the angle between the two vectors, and the directional consistency is 0.897.

[0077] Simultaneously, a vibration safety factor is calculated based on the normalized vibration vector to reflect the absolute vibration intensity level. The vibration safety factor = 1 - the mean of each dimension of the normalized vibration vector, where the mean is the arithmetic mean of the normalized vibration amplitudes in each preset direction. The closer the normalized vibration amplitude is to 1, the closer the vibration is to the historical maximum value, and the lower the intensity factor; conversely, when the overall vibration amplitude is small, the intensity factor tends to be close to 1. If the current value exceeds the range defined by the historical minimum and maximum values, the normalized value is truncated to 0 or 1.

[0078] The normalized vibration amplitude mentioned above is obtained by linear transformation of the historical minimum and maximum vibration amplitude values ​​stored in the polishing vibration database. These minimum and maximum values ​​are fixed historical benchmarks and do not change with the current parameters to be verified.

[0079] For example, to calculate the vibration safety factor: the mean of each dimension of the normalized vibration vector = (0.333 + 0.259 + 0.185 + 0.259) / 4 = 0.259, and the vibration safety factor = 1 - 0.259 = 0.741.

[0080] Finally, the vibration acceptability is obtained by multiplying directional consistency by the vibration safety factor. Vibration acceptability = directional consistency × (1 - mean of each dimension of the normalized vibration vector). This definition allows vibration acceptability to simultaneously reflect two dimensions: whether the vibration distribution pattern matches the roughness distribution pattern and whether the absolute vibration intensity is safe. Vibration acceptability approaches 1 only when the vibration distribution matches the roughness distribution shape and the absolute vibration intensity is within a safe range. When the vibration amplitude in all directions is synchronously large, even if the directional consistency remains the same, the decrease in the vibration safety factor will reduce vibration acceptability, preventing the scale invariance of cosine similarity from masking the risk of excessive vibration.

[0081] The calculated vibration acceptance rate is used as the output result. If the vibration acceptance rate is higher than the preset acceptance threshold, the current parameters pass the vibration verification and can proceed to the next optimization step; if it is lower than the threshold, the initial polishing parameters need to be adjusted and vibration prediction and verification need to be performed again.

[0082] The preset acceptance threshold is usually determined based on process experience or the statistical distribution of vibration acceptability corresponding to qualified polishing batches in historical data. Generally, the lower quartile of vibration acceptability of historical qualified batches is taken as the threshold.

[0083] For example, if the vibration acceptance rate of 50 qualified batches is statistically analyzed and the lower quartile value is 0.60, then the acceptance threshold is set to 0.60. A value below the acceptance threshold indicates an abnormal correlation between vibration and roughness distribution, or that the absolute vibration intensity is too high, which may indicate equipment failure or parameter mismatch, requiring investigation or adjustment.

[0084] For example, vibration acceptance level = directional consistency × vibration safety factor = 0.897 × 0.741 ≈ 0.665. The current vibration acceptance level of 0.665 > 0.60, and after verification, the final output is 0.665 as the vibration acceptance level.

[0085] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0086] In summary, this step assesses the feasibility of vibration analysis using comprehensive indicators before processing. It examines both the matching degree between vibration distribution and roughness distribution, as well as whether the absolute intensity of vibration is safe. This process screens out parameter combinations that may cause abnormal vibration or uneven polishing, providing a safety boundary for subsequent optimization and avoiding ineffective trial polishing and equipment damage.

[0087] S400: Based on the initial polishing roughness array and vibration acceptance, the polishing parameter optimization step size configuration is performed to obtain the polishing parameter optimization step size array. The optimization is then adjusted and iterated to obtain the optimal polishing parameters.

[0088] like Figure 2 As shown, this step first obtains the allowable roughness threshold and calculates the fitness of the initial polishing parameters as the optimization benchmark. Then, based on the deviation of the predicted roughness from the threshold at each detection point, the deviation step size coefficient is configured. Combined with the vibration acceptance, the vibration step size adjustment coefficient is configured. The preset adjustment step size is double-corrected to obtain the optimized step size array for adjusting the polishing parameters. Finally, the initial polishing parameters are adjusted in multiple directions and the fitness is iteratively evaluated. After convergence, the optimal polishing parameters are output.

[0089] Step S400 in the method provided by the present invention includes: Obtain the acceptable roughness threshold for polishing; Based on the initial polishing roughness array, the initial polishing vibration parameter array, and the vibration acceptance, the initial polishing fitness of the initial polishing parameters is calculated. The deviation of each initial polishing roughness in the initial polishing roughness array from the allowable roughness threshold is calculated to obtain multiple initial polishing deviations; Obtain the preset adjustment step size for adjusting the polishing parameters, and use multiple initial polishing deviations to adjust and calculate the preset adjustment step size to obtain an optimized step size array for polishing parameters; Based on the vibration acceptance, a vibration step size adjustment coefficient is configured, and the polishing parameter optimization step size array is adjusted and calculated to obtain the adjusted polishing parameter optimization step size array. The initial polishing parameters are adjusted by using multiple adjustment polishing parameter optimization step sizes within the adjustment polishing parameter optimization step size array to obtain multiple adjustment polishing parameters, and then iterative optimization and polishing fitness calculation are performed. After optimization and convergence, the optimal polishing parameters with the greatest polishing fitness are obtained.

[0090] In this step, the first step is to obtain the allowable roughness threshold required by the polishing process, which is the upper limit of the roughness of the inner wall that must be achieved after polishing. The allowable roughness threshold is set according to the product technical specifications or industry standards.

[0091] For example, the stainless steel tube processed in the previous example is used in the semiconductor industry. According to the SEMI standard, the inner wall roughness requirement is Ra≤0.2μm, so the allowable roughness threshold is set to 0.2μm. If the predicted roughness is ≤0.2μm, it meets the standard; if it is >0.2μm, the step size corresponding to that detection point needs to be increased in the optimization process to prioritize improvement.

[0092] Further, based on the initial polishing roughness array, the initial polishing vibration parameter array, and the vibration acceptance, the initial polishing fitness of the initial polishing parameters is calculated, including: Calculate the mean values ​​of the initial polishing roughness array and the initial polishing vibration parameter array to obtain the mean value of the initial polishing roughness and the mean value of the initial polishing vibration parameters; The compliance margin between the initial average polishing roughness and the allowable roughness threshold is calculated to obtain the basic roughness adaptability. Calculate the threshold of polishing vibration parameters and the compliance margin of the mean of initial polishing vibration parameters, and use the vibration acceptability correction to obtain the basic vibration fitness. The initial polishing fitness is calculated based on the basic roughness fitness and basic vibration fitness.

[0093] Specifically, the arithmetic mean of all predicted values ​​in the initial polishing roughness array is first calculated to obtain the mean initial polishing roughness; the arithmetic mean of the vibration amplitudes in each direction in the initial polishing vibration parameter array is then calculated to obtain the mean initial polishing vibration parameter.

[0094] For example, the summation and averaging of the four predicted values ​​in the initial polishing roughness array [0.12, 0.14, 0.16, 0.13] yields an initial polishing roughness mean of 0.14 μm. The initial polishing vibration parameter array [0.12, 0.10, 0.08, 0.10] has a mean of (0.12 + 0.10 + 0.08 + 0.10) / 4 = 0.10 mm, meaning the initial polishing vibration parameter mean is 0.10 mm.

[0095] Secondly, the compliance margin between the initial average polishing roughness and the allowable roughness threshold is calculated to obtain the basic roughness fitness. Basic roughness fitness = 1 - initial average polishing roughness / allowable roughness threshold, with a value in the range [0,1]. The closer to 1, the better the polishing effect is than required; the closer to 0, the closer to or exceeding the threshold.

[0096] For example, with an allowable roughness threshold of 0.2 μm and an initial average polishing roughness of 0.14 μm, the basic roughness adaptability = 1 - 0.14 / 0.2 = 1 - 0.7 = 0.3, indicating that the current polishing effect is better than required.

[0097] Next, a preset polishing vibration parameter threshold is obtained, and the compliance margin between the polishing vibration parameter threshold and the initial average polishing vibration parameter is calculated. Then, the compliance margin is corrected using vibration acceptability to obtain the basic vibration fitness. The correction method is to multiply the compliance margin by the vibration acceptability, so that the coordination between the vibration state and the roughness distribution affects the basic vibration fitness.

[0098] The polishing vibration parameter threshold refers to the maximum vibration parameter that will not cause polishing abnormalities in historical polishing processes. It is taken from the maximum vibration amplitude in each direction corresponding to all qualified batches in the polishing vibration database. A qualified batch is defined as a processing record that meets the surface roughness standard after polishing, without over-polishing or under-polishing, and without equipment alarms. The average of the maximum vibration amplitude is used as the vibration parameter threshold, which can be dynamically updated as data accumulates.

[0099] For example, in the polishing vibration database, there are 85 qualified batches. The maximum vibration amplitudes in the four circumferential directions are 0.28mm, 0.22mm, 0.18mm, and 0.22mm, respectively. The average value is taken as (0.28+0.22+0.18+0.22) / 4=0.225mm, approximately 0.23mm, which is used as the average threshold value for the polishing vibration parameter. The basic vibration adaptability is calculated as (1-0.10 / 0.23)×0.665≈0.376.

[0100] Finally, the initial polishing fitness is calculated based on the base roughness fitness and the base vibration fitness. The calculation method involves multiplying the two, so that the performance of both dimensions collectively determines the final fitness; poor performance in either dimension will significantly lower the overall fitness. For example, if the base vibration fitness is 0.376 and the base roughness fitness is 0.3, the initial polishing fitness = 0.3 × 0.376 ≈ 0.113.

[0101] Furthermore, the deviation of each predicted roughness in the initial polishing roughness array from the allowable roughness threshold is calculated to obtain multiple initial polishing deviations. The deviation reflects the difference between the predicted value at each detection point and the allowable upper limit; a positive value indicates that the standard is not exceeded but there is a margin, while a negative value indicates that the standard is exceeded and improvement is needed.

[0102] For example, with an allowable roughness threshold of 0.2 μm, the initial polishing roughness array is [0.12, 0.14, 0.16, 0.13], and the deviations are [0.08, 0.06, 0.04, 0.07], in μm.

[0103] Furthermore, the preset adjustment step size corresponding to each polishing parameter is obtained, and the adjustment step size is the basic adjustment range of each parameter in each parameter iteration. The preset adjustment step size is weighted and adjusted using each initial polishing deviation, thereby obtaining a set of polishing parameter optimization step sizes corresponding to each detection point, which are summarized into a polishing parameter optimization step size array.

[0104] The preset adjustment step size is typically 5% to 10% of the feasible range of each parameter. For example, for a polishing speed range of 1000–1900 rpm (900 rpm increments), 5% (45 rpm) is taken and rounded up to 50 rpm as the preset speed step size; for a pressure range of 0.3–0.7 MPa (0.4 MPa increments), 5% (0.02 MPa) is taken as the preset pressure step size; for a concentration range of 20%–50% (30% increments), 5% (1.5%) is taken and rounded up to 1% as the preset concentration step size; and for a time range of 10–30 minutes (20 minutes increments), 5% (1 minute) is taken as the preset time step size. The final preset adjustment step size is: speed 50 rpm, pressure 0.02 MPa, concentration 1%, and time 1 minute.

[0105] The method for adjusting the preset adjustment step size using each initial polishing deviation is as follows: Set the deviation step size coefficient α, and set it in segments according to the initial polishing deviation: when the initial polishing deviation ≤ 0, α is 2.5 to accelerate correction with the maximum step size; when 0 < initial polishing deviation ≤ 0.03μm, α is 2.0; when 0.03 < initial polishing deviation ≤ 0.06μm, α is 1.5; when 0.06 < initial polishing deviation ≤ 0.10μm, α is 1.0; when the initial polishing deviation > 0.10μm, α is 0.5 to reduce the step size and avoid over-adjustment.

[0106] For example, the preset step size is 50 rpm, 0.02 MPa, 1% concentration, and 1 minute. The deviations of the four detection points are [0.08, 0.06, 0.04, 0.07] respectively. The first point, 0.08 μm, corresponds to α=1.0, with a step size of [50 rpm, 0.02 MPa, 1%, 1 minute]; the second point, 0.06 μm, corresponds to α=1.5, with a step size of [75 rpm, 0.03 MPa, 1.5%, 1.5 minutes]; the third point, 0.04 μm, corresponds to α=1.5, with a step size of [75 rpm, 0.03 MPa, 1.5%, 1.5 minutes]; and the fourth point, 0.07 μm, corresponds to α=1.0, with a step size of [50 rpm, 0.02 MPa, 1%, 1 minute]. This constitutes four sets of polishing parameter optimization step size arrays.

[0107] Furthermore, a vibration step size adjustment coefficient β is configured based on the vibration acceptability. A higher vibration acceptability indicates a more harmonious relationship between the current vibration state and the roughness distribution, allowing the vibration step size adjustment amplitude to remain normal. Conversely, a lower vibration acceptability necessitates increasing the step size to accelerate the escape from the current parameter range or adjusting the parameter direction to improve the vibration state. The vibration step size adjustment coefficient is then used to uniformly adjust the step sizes of each group in the polishing parameter optimization step size array, resulting in the adjusted polishing parameter optimization step size array.

[0108] The vibration step size adjustment coefficient β is set in segments according to the vibration acceptance level: when the vibration acceptance level is ≥0.85, it indicates that the vibration condition is good, β=1.0, and the step size remains unchanged; when 0.70≤vibration acceptance level<0.85, it indicates that the vibration condition is average, β=1.3, and the step size is moderately increased; when 0.55≤vibration acceptance level<0.70, it indicates that the vibration condition is poor, β=1.6, and the step size is significantly increased; when the vibration acceptance level<0.55, it indicates that the vibration condition is unacceptable, β=2.0, the step size is further increased, and it is recommended to check for equipment abnormalities.

[0109] Among them, the aforementioned segmentation thresholds of 0.85, 0.70, and 0.55 are cosine similarity evaluation thresholds for directional consistency. Their correspondence with the comprehensive vibration acceptability value is as follows: above 0.85 indicates high similarity, 0.70–0.85 indicates moderate similarity, 0.55–0.70 indicates low similarity, and below 0.55 indicates weak similarity or irrelevance. Vibration acceptability integrates directional consistency and vibration safety factor; its absolute value is lower than directional consistency. Therefore, the applicability of the segmentation thresholds is judged based on directional consistency, and the vibration step size adjustment coefficient is configured according to the interval in which the vibration acceptability falls.

[0110] For example, if the current vibration acceptance is 0.665, which falls within the range of 0.55 ≤ vibration acceptance < 0.70, it indicates a poor vibration condition. The vibration step size adjustment coefficient is set to β = 1.6. After the adjustment at the second point, the step size is changed from [75 rpm, 0.03 MPa, 1.5%, 1.5 minutes] to [120 rpm, 0.048 MPa, 2.4%, 2.4 minutes]. The other points are similarly amplified by β = 1.6.

[0111] Furthermore, the step size of each group within the step size array is optimized by adjusting the polishing parameters. The initial polishing parameters are adjusted accordingly to obtain multiple adjusted polishing parameters. The adjustment method involves adding or subtracting each component of the initial polishing parameters from the corresponding step size, depending on the optimization direction. Each group of step sizes corresponds to a set of adjusted polishing parameters, resulting in multiple sets of adjusted polishing parameters equal to the number of detection points.

[0112] For example, the initial polishing parameters are [rotation speed 1765 rpm, pressure 0.64 MPa, concentration 45.5%, time 27 minutes]. Taking the amplified adjustment step size corresponding to the second point [120 rpm, 0.048 MPa, 2.4%, 2.4 minutes], and increasing it, we obtain the adjusted polishing parameters [1885 rpm, 0.688 MPa, 47.9%, 29.4 minutes]. Taking the amplified adjustment step size corresponding to the first point [80 rpm, 0.032 MPa, 1.6%, 1.6 minutes], and increasing it, we obtain [1845 rpm, 0.672 MPa, 47.1%, 28.6 minutes]. The third and fourth points are generated similarly. This process generates four sets of adjusted polishing parameters.

[0113] Furthermore, each set of adjusted polishing parameters is input into the polishing prediction model and the polishing vibration database, respectively, to re-predict the polishing roughness array and vibration parameter array, calculate the vibration acceptance, and then calculate the polishing fitness of each set of polishing parameters according to the aforementioned fitness formula. Multiple sets of adjusted polishing parameters are used as the initial particle swarm in the particle swarm optimization algorithm. The polishing fitness of each set of adjusted polishing parameters is used as the objective function value of the corresponding particle. The particle swarm optimization algorithm is used to update the particle velocity and position. After multiple iterations, when the convergence condition is met, the polishing parameters corresponding to the particle with the highest global fitness are output as the optimal polishing parameters.

[0114] It should be noted that during the iterative optimization process, if the adjusted parameter exceeds the feasible upper and lower limits of the polishing parameter space, the parameter will be truncated to the nearest boundary value.

[0115] For example, in the first iteration, the initial polishing parameters were [rotation speed 1765 rpm, pressure 0.64 MPa, concentration 45.5%, time 27 minutes], vibration acceptability 0.665, and vibration step size adjustment coefficient β=1.6. The deviation of the second detection point was 0.06 μm, corresponding to a deviation step size coefficient α=1.5. Its adjustment step size = [75 rpm, 0.03 MPa, 1.5%, 1.5 minutes] × 1.6 = [120 rpm, 0.048 MPa, 2.4%, 2.4 minutes]. The adjusted parameters were [1885 rpm, 0.688 MPa, 47.9%, 29.4 minutes]. Based on prediction and fitness calculation, the fitness of each group's adjusted polishing parameters is used as the initial fitness value of the corresponding particle in the particle swarm optimization algorithm. According to the particle swarm optimization algorithm update rule, the parameter corresponding to the second detection point [1885rpm, 0.688MPa, 47.9%, 29.4 minutes] is taken as the current global best particle with a fitness of 0.32.

[0116] The second iteration used the globally optimal particle from the first round [1885 rpm, 0.688 MPa, 47.9%, 29.4 minutes] as the baseline. After updating the particle positions according to the particle swarm optimization algorithm, the point with the smallest deviation became the third detection point, with a deviation of 0.02 μm, corresponding to α=2.0; the vibration acceptance decreased to 0.58, corresponding to β=1.6. The step size was adjusted to [50 rpm, 0.02 MPa, 1%, 1 minute] × 2.0 × 1.6 = [160 rpm, 0.064 MPa, 3.2%, 3.2 minutes], and the adjusted parameters were [2045 rpm, 0.752 MPa, 51.1%, 32.6 minutes]. The following parameters were selected: a rotation speed of 2045 rpm, exceeding the feasible upper limit of 1900 rpm; a pressure of 0.752 MPa, exceeding the upper limit of 0.7 MPa; a concentration of 51.1%, exceeding the upper limit of 50%; and a time of 32.6 minutes, exceeding the upper limit of 30 minutes. All of these were truncated to boundary values, resulting in [1900 rpm, 0.7 MPa, 50%, 30 minutes]. The fitness of this set of parameters was 0.41, representing the optimal result for the second round.

[0117] The third iteration used the globally optimal particle [1900 rpm, 0.7 MPa, 50%, 30 minutes] from the second round as a baseline. After updating the particle positions according to the particle swarm optimization algorithm, the point with the largest deviation became the first detection point, with a deviation of 0.07 μm, corresponding to α=1.0; the vibration acceptability recovered to 0.72, corresponding to β=1.3. The adjustment step size was [50 rpm, 0.02 MPa, 1%, 1 minute] × 1.0 × 1.3 = [65 rpm, 0.026 MPa, 1.3%, 1.3 minutes]. Since the rotation speed was already at the upper limit of the feasible space, the rotation speed was not increased further. The pressure, concentration, and time were adjusted, and the rotation speed was appropriately reduced to expand the search range. The adjusted parameters were [1850 rpm, 0.7 MPa, 50%, 30 minutes]. The fitness of this set of parameters was 0.45, which was optimal for the third round.

[0118] The fourth iteration uses the globally optimal particle [1850 rpm, 0.7 MPa, 50%, 30 minutes] from the third iteration as a baseline. After updating the particle positions according to the particle swarm optimization algorithm, the point with the largest deviation becomes the fourth detection point, with a deviation of 0.05 μm, corresponding to α=1.5; the vibration acceptance is 0.75, corresponding to β=1.3. The adjustment step size is [50 rpm, 0.02 MPa, 1%, 1 minute] × 1.5 × 1.3 = [97.5 rpm, 0.039 MPa, 1.95%, 1.95 minutes], which is rounded to [100 rpm, 0.04 MPa, 2%, 2 minutes]. The adjusted parameters are [1950 rpm, 0.74 MPa, 52%, 32 minutes], where the rotation speed, pressure, concentration, and time all exceed the feasible space boundary, and are truncated to [1900 rpm, 0.7 MPa, 50%, 30 minutes]. The fitness of this set of parameters is 0.46, which is the best in the 4th round.

[0119] Finally, when the polishing fitness increment between two adjacent iterations is lower than the preset convergence threshold, or when the global optimal fitness in the particle swarm optimization algorithm shows no significant improvement for several consecutive iterations, convergence is determined and the iteration stops. The set of polishing parameters with the highest polishing fitness in the last iteration before convergence is output as the optimal polishing parameters.

[0120] The convergence threshold is the lower limit of the fitness increment at which the optimization iteration is terminated. The optimization is considered to have converged when the fitness increment of two adjacent iterations is less than the convergence threshold. The convergence threshold is typically set to 1%–2% of the initial fitness, or a fixed value depending on the required process precision. Since the fitness is limited to the [0,1] interval, a fixed value of 0.01 or 2% of the initial fitness can be used for the convergence threshold. Setting it too high will lead to premature convergence and failure to reach the optimal state; setting it too low will increase the number of iterations and computation time.

[0121] The maximum number of iterations is estimated based on the ratio of the parameter space size to the step size, typically using the average number of selectable values ​​for each dimension of the parameter space as a reference. For example, with a polishing speed range of 1000–1900 rpm and a step size of 50 rpm, there are approximately 18 selectable values; with a pressure range of 0.3–0.7 MPa and a step size of 0.02 MPa, there are approximately 20 selectable values. With an average of approximately 20 selectable values, setting the maximum number of iterations to 20–30 is sufficient to cover most of the search path.

[0122] For example, the initial fitness is 0.113, and a fixed value of 0.01 is used as the convergence threshold. In the above iteration process, the fitness is 0.45 in the 3rd round and 0.46 in the 4th round, with an increment of 0.01, reaching the convergence threshold of 0.01, and the iteration stops. Finally, the polishing parameters corresponding to the globally optimal particle after the 4th iteration are [1900 rpm, 0.7 MPa, 50%, 30 minutes]. Under the premise of meeting the allowable roughness threshold, the vibration state and roughness distribution have the best coordination.

[0123] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0124] In summary, this step integrates the requirements for roughness compliance and vibration coordination into a variable step-size search framework. This allows the optimization process to automatically focus on parameters with large deviations and high vibration risks for accelerated correction, balancing search efficiency and convergence accuracy. Ultimately, it obtains the optimal process parameters for vibration under the premise of meeting the allowable roughness threshold, eliminating reliance on manual experience and the cost of repeated trial and error. This ensures the uniformity of circumferential polishing of the entire inner wall of the pipe and, under multi-section detection conditions, also considers the uniformity of polishing along the entire length of the pipe.

[0125] In summary, this invention achieves fully automated decision-making of the stainless steel pipe inner wall polishing process parameters from initial configuration to final optimization, significantly shortening the debugging cycle, reducing trial polishing costs, and ensuring the polishing uniformity and processing stability of the entire pipe inner wall; under multi-section detection conditions, it can also take into account the polishing uniformity along the entire pipe length, thereby improving the overall processing stability.

[0126] Example 2, as Figure 3 As shown, the present invention provides a system for optimizing the polishing process of the inner wall of stainless steel pipes, the system comprising: The data acquisition and initial configuration module 11 is used to perform sampling endoscopic inspection on stainless steel pipes to obtain an inner wall roughness array and configure the initial polishing parameters for the polishing process.

[0127] The process involves sampling and endoscopic inspection of stainless steel pipes to obtain an inner wall roughness array, and configuring initial polishing parameters for the polishing process, including: Randomly sample stainless steel pipes from the same batch, and select test points in multiple preset directions within the sampled stainless steel pipes. Use industrial endoscopic equipment to perform internal roughness testing to obtain an internal roughness array. Obtain the polishing parameter space and select to generate initial polishing parameters, including the polishing speed.

[0128] The prediction and simulation module 12 is used to perform polishing prediction based on the initial polishing parameters and the inner wall roughness array to obtain the initial polishing roughness array, and to perform vibration prediction to obtain the initial polishing vibration parameter array.

[0129] Specifically, based on the initial polishing parameters and the inner wall roughness array, polishing prediction is performed to obtain the initial polishing roughness array, including: Based on the stainless steel pipe inner wall polishing test records, the sample polishing parameter set and the sample original inner wall roughness set were collected, as well as the inner wall roughness after polishing according to different sample polishing parameters and sample original inner wall roughness, to obtain the sample polishing roughness set. A model training and testing mechanism is adopted to divide the sample polishing parameter set, the sample original inner wall roughness set, and the sample polishing roughness set to obtain training data and test data. A qualified polishing prediction model is obtained by training and testing. The initial polishing parameters and each inner wall roughness in the inner wall roughness array are input into the polishing prediction model, and the initial polishing roughness array is output.

[0130] Among them, vibration prediction is performed to obtain an initial polishing vibration parameter array, including: Access the polishing vibration database; The initial polishing parameters are used to search the polishing vibration database to obtain an initial polishing vibration parameter array, wherein the initial polishing vibration parameter array includes vibration amplitudes in multiple preset directions.

[0131] The steps for constructing the polishing vibration database include: Collect a set of sample polishing parameters and test the vibration amplitude of the polishing equipment in multiple preset directions under different sample polishing parameters to obtain a set of sample vibration amplitude arrays; A mapping relationship is constructed between the sample polishing parameter set and the sample vibration amplitude array set, and fuzzy mapping rules are configured to build a polishing vibration database. The fuzzy mapping rules include matching the closest sample polishing parameters to the input polishing parameters for mapping.

[0132] The vibration acceptability verification module 13 is used to verify the vibration acceptability of the initial polishing vibration parameter array and the inner wall roughness array to obtain the vibration acceptability.

[0133] The vibration acceptability verification of the initial polishing vibration parameter array and the inner wall roughness array is performed to obtain the vibration acceptability, including: The inner wall roughness array is normalized to obtain a normalized roughness array; The initial polishing vibration parameter array is normalized to obtain a normalized vibration parameter array; The vibration acceptability was calculated and verified by the normalized roughness array and the normalized vibration parameter array to obtain the vibration acceptability.

[0134] The vibration acceptability calculation and verification are performed on the normalized roughness array and the normalized vibration parameter array to obtain the vibration acceptability, including: Based on the normalized roughness array and the normalized vibration parameter array, a normalized roughness vector and a normalized vibration vector are constructed. Calculate the similarity between the normalized roughness vector and the normalized vibration vector to obtain directional consistency; Calculate the vibration safety factor based on the normalized vibration vector; Multiplying the directional consistency by the vibration safety factor yields the vibration acceptability.

[0135] The optimization iteration and output module 14 is used to configure the polishing parameter optimization step size according to the initial polishing roughness array and vibration acceptance, obtain the polishing parameter optimization step size array, and perform adjustment and optimization iteration to obtain the optimal polishing parameters.

[0136] Specifically, based on the initial polishing roughness array and vibration tolerance, a polishing parameter optimization step size configuration is performed to obtain a polishing parameter optimization step size array. This array is then adjusted and iterated to obtain the optimal polishing parameters, including: Obtain the acceptable roughness threshold for polishing; Based on the initial polishing roughness array, the initial polishing vibration parameter array, and the vibration acceptance, the initial polishing fitness of the initial polishing parameters is calculated. The deviation of each initial polishing roughness in the initial polishing roughness array from the allowable roughness threshold is calculated to obtain multiple initial polishing deviations; Obtain the preset adjustment step size for adjusting the polishing parameters, and use multiple initial polishing deviations to adjust and calculate the preset adjustment step size to obtain an optimized step size array for polishing parameters; Based on the vibration acceptance, a vibration step size adjustment coefficient is configured, and the polishing parameter optimization step size array is adjusted and calculated to obtain the adjusted polishing parameter optimization step size array. The initial polishing parameters are adjusted by using multiple adjustment polishing parameter optimization step sizes within the adjustment polishing parameter optimization step size array to obtain multiple adjustment polishing parameters, and then iterative optimization and polishing fitness calculation are performed. After optimization and convergence, the optimal polishing parameters with the greatest polishing fitness are obtained.

[0137] The calculation of the initial polishing fitness of the initial polishing parameters based on the initial polishing roughness array, the initial polishing vibration parameter array, and the vibration acceptability includes: Calculate the mean values ​​of the initial polishing roughness array and the initial polishing vibration parameter array to obtain the mean value of the initial polishing roughness and the mean value of the initial polishing vibration parameters; The compliance margin between the initial average polishing roughness and the allowable roughness threshold is calculated to obtain the basic roughness adaptability. Calculate the threshold of polishing vibration parameters and the compliance margin of the mean of initial polishing vibration parameters, and use the vibration acceptability correction to obtain the basic vibration fitness. The initial polishing fitness is calculated based on the basic roughness fitness and basic vibration fitness.

[0138] 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.

[0139] This specification and accompanying drawings are merely illustrative examples of the invention and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its scope. Therefore, if such modifications and modifications fall within the scope of the invention and its equivalents, the invention is intended to include these modifications and modifications.

Claims

1. An optimized method for polishing the inner wall of stainless steel pipes, characterized in that, The method includes: Sampling endoscopic inspection was performed on the stainless steel pipe to obtain the inner wall roughness array, and the initial polishing parameters for the polishing process were configured. Based on the initial polishing parameters and the inner wall roughness array, polishing prediction is performed to obtain the initial polishing roughness array, and vibration prediction is performed to obtain the initial polishing vibration parameter array. The vibration acceptability of the initial polishing vibration parameter array and the inner wall roughness array is verified to obtain the vibration acceptability. Based on the initial polishing roughness array and vibration tolerance, the polishing parameter optimization step size configuration is performed to obtain the polishing parameter optimization step size array. The optimization is then adjusted and iterated to obtain the optimal polishing parameters.

2. The optimized polishing process for the inner wall of stainless steel pipe according to claim 1, characterized in that, Sampling endoscopic inspection was performed on the stainless steel pipe to obtain an inner wall roughness array, and initial polishing parameters for the polishing process were configured, including: Randomly sample stainless steel pipes from the same batch, and select test points in multiple preset directions within the sampled stainless steel pipes. Use industrial endoscopic equipment to perform internal roughness testing to obtain an internal roughness array. Obtain the polishing parameter space and select to generate initial polishing parameters, including the polishing speed.

3. The optimized polishing process for the inner wall of stainless steel pipe according to claim 1, characterized in that, Based on the initial polishing parameters and the inner wall roughness array, polishing prediction is performed to obtain the initial polishing roughness array, including: Based on the stainless steel pipe inner wall polishing test records, the sample polishing parameter set and the sample original inner wall roughness set were collected, as well as the inner wall roughness after polishing according to different sample polishing parameters and sample original inner wall roughness, to obtain the sample polishing roughness set. A model training and testing mechanism is adopted to divide the sample polishing parameter set, the sample original inner wall roughness set, and the sample polishing roughness set to obtain training data and test data. A qualified polishing prediction model is obtained by training and testing. The initial polishing parameters and each inner wall roughness in the inner wall roughness array are input into the polishing prediction model, and the initial polishing roughness array is output.

4. The optimized polishing process for the inner wall of stainless steel pipe according to claim 1, characterized in that, Vibration prediction is performed to obtain an initial polishing vibration parameter array, including: Access the polishing vibration database; The initial polishing parameters are used to search the polishing vibration database to obtain an initial polishing vibration parameter array, wherein the initial polishing vibration parameter array includes vibration amplitudes in multiple preset directions.

5. The optimized polishing process for the inner wall of stainless steel pipe according to claim 4, characterized in that, The steps for constructing the polishing vibration database include: Collect a set of sample polishing parameters and test the vibration amplitude of the polishing equipment in multiple preset directions under different sample polishing parameters to obtain a set of sample vibration amplitude arrays; A mapping relationship is constructed between the sample polishing parameter set and the sample vibration amplitude array set, and fuzzy mapping rules are configured to build a polishing vibration database. The fuzzy mapping rules include matching the closest sample polishing parameters to the input polishing parameters for mapping.

6. The optimized polishing process for the inner wall of stainless steel pipe according to claim 1, characterized in that, The initial polishing vibration parameter array and the inner wall roughness array are subjected to vibration acceptability verification to obtain vibration acceptability, including: The inner wall roughness array is normalized to obtain a normalized roughness array; The initial polishing vibration parameter array is normalized to obtain a normalized vibration parameter array; The vibration acceptability was calculated and verified by the normalized roughness array and the normalized vibration parameter array to obtain the vibration acceptability.

7. The optimized polishing process for the inner wall of stainless steel pipe according to claim 6, characterized in that, The vibration acceptability of the normalized roughness array and the normalized vibration parameter array is calculated and verified to obtain the vibration acceptability, including: Based on the normalized roughness array and the normalized vibration parameter array, a normalized roughness vector and a normalized vibration vector are constructed. Calculate the similarity between the normalized roughness vector and the normalized vibration vector to obtain directional consistency; Calculate the vibration safety factor based on the normalized vibration vector; Multiplying the directional consistency by the vibration safety factor yields the vibration acceptability.

8. The optimized polishing process for the inner wall of stainless steel pipe according to claim 1, characterized in that, Based on the initial polishing roughness array and vibration tolerance, the polishing parameter optimization step size is configured to obtain the polishing parameter optimization step size array. This array is then adjusted and iterated to obtain the optimal polishing parameters, including: Obtain the acceptable roughness threshold for polishing; Based on the initial polishing roughness array, the initial polishing vibration parameter array, and the vibration acceptance, the initial polishing fitness of the initial polishing parameters is calculated. The deviation of each initial polishing roughness in the initial polishing roughness array from the allowable roughness threshold is calculated to obtain multiple initial polishing deviations; Obtain the preset adjustment step size for adjusting the polishing parameters, and use multiple initial polishing deviations to adjust and calculate the preset adjustment step size to obtain an optimized step size array for polishing parameters; Based on the vibration acceptance, a vibration step size adjustment coefficient is configured, and the polishing parameter optimization step size array is adjusted and calculated to obtain the adjusted polishing parameter optimization step size array. The initial polishing parameters are adjusted by using multiple adjustment polishing parameter optimization step sizes within the adjustment polishing parameter optimization step size array to obtain multiple adjustment polishing parameters, and then iterative optimization and polishing fitness calculation are performed. After optimization and convergence, the optimal polishing parameters with the greatest polishing fitness are obtained.

9. The optimized polishing process for the inner wall of a stainless steel pipe according to claim 8, characterized in that, Based on the initial polishing roughness array, the initial polishing vibration parameter array, and the vibration acceptability, the initial polishing fitness of the initial polishing parameters is calculated, including: Calculate the mean values ​​of the initial polishing roughness array and the initial polishing vibration parameter array to obtain the mean value of the initial polishing roughness and the mean value of the initial polishing vibration parameters; The compliance margin between the initial average polishing roughness and the allowable roughness threshold is calculated to obtain the basic roughness adaptability. Calculate the threshold of polishing vibration parameters and the compliance margin of the mean of initial polishing vibration parameters, and use the vibration acceptability correction to obtain the basic vibration fitness. The initial polishing fitness is calculated based on the basic roughness fitness and basic vibration fitness.

10. A system for optimizing the polishing process of the inner wall of stainless steel pipes, characterized in that, The system is used to perform the stainless steel pipe inner wall polishing process optimization method according to any one of claims 1 to 9, the system comprising: The data acquisition and initial configuration module is used to perform sampling endoscopic inspection on stainless steel pipes to obtain an inner wall roughness array and configure the initial polishing parameters for the polishing process. The prediction and simulation module is used to perform polishing prediction based on the initial polishing parameters and the inner wall roughness array to obtain the initial polishing roughness array, and to perform vibration prediction to obtain the initial polishing vibration parameter array. The vibration acceptability verification module is used to verify the vibration acceptability of the initial polishing vibration parameter array and the inner wall roughness array to obtain the vibration acceptability. The optimization iteration and output module is used to configure the polishing parameter optimization step size according to the initial polishing roughness array and vibration acceptance, obtain the polishing parameter optimization step size array, and perform adjustment and optimization iteration to obtain the optimal polishing parameters.

Images