Polishing process optimization method and system combined with big data analysis

Abstract

The embodiment of the invention provides a polishing process optimization method and system combined with big data analysis, and relates to the technical field of big data, first, workpiece basic attributes, polishing equipment operation, process parameter execution and polishing effect feedback data are integrated, and a polishing process multi-source data linkage set is established; based on the polishing process multi-source data linkage set, mining dynamic coupling relations of different process parameters to form a coupling rule set; according to the technological parameter coupling rule set and historical polishing effect feedback data, after a prediction model with workpiece basic attributes and initial technological parameters as input and predicted polishing effects as output is constructed, basic attribute data of a to-be-polished workpiece is input, the technological parameters are iteratively adjusted in combination with real-time equipment operation data, a dynamic optimization parameter sequence is generated, and the polishing effect is predicted. And finally, outputting a complete optimization scheme integrating equipment operation, a process time sequence and a quality control node. According to the method, data are comprehensively utilized, the polishing process is accurately optimized, and quality and efficiency are improved.

Description

Technical Field

[0001] This application relates to the field of big data technology, and more specifically, to a polishing process optimization method and system that combines big data analysis. Background Technology

[0002] In the field of polishing technology, as industrial production demands increasingly higher product quality and efficiency, traditional polishing process optimization methods have gradually revealed numerous limitations. Currently, many polishing process optimizations rely heavily on experienced operators who adjust process parameters based on their accumulated experience to achieve the desired polishing effect. However, this reliance on manual experience has significant shortcomings. On the one hand, manual experience is subjective and limited; different operators may have different understandings and adjustment methods for process parameters, making it difficult to guarantee the stability and consistency of polishing results. On the other hand, with the increasing variety and complexity of products, relying solely on manual experience makes it difficult to comprehensively consider the impact of various factors on polishing results, hindering the achievement of refined process optimization.

[0003] Furthermore, while some existing polishing process optimization methods do collect some data, the data sources are relatively singular. They typically focus only on the process parameters themselves, neglecting the comprehensive utilization of information from multiple aspects such as workpiece basic properties, polishing equipment operating status, and polishing effect feedback. Such optimization methods relying on a single data source cannot comprehensively and accurately reflect the complex relationships in the polishing process, making it difficult to uncover the potential dynamic coupling patterns between process parameters. This limits the improvement of polishing process optimization effects and fails to meet the demands of modern industrial production for high-quality, high-efficiency polishing processes. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a polishing process optimization method and system that combines big data analysis.

[0005] According to a first aspect of this application, a polishing process optimization method incorporating big data analysis is provided, the method comprising: By combining workpiece basic attribute data, polishing equipment operation data, process parameter execution data, and polishing effect feedback data, a multi-source data linkage set for polishing process is established. The multi-source data linkage set for polishing process includes real-time synchronization links and historical association indexes for each data type. Based on the multi-source data linkage set of the polishing process, the dynamic coupling relationship between different process parameters is mined to form a process parameter coupling rule set, which includes parameter change coordination logic and parameter adaptation constraints. Based on the process parameter coupling rule set and historical polishing effect feedback data, a polishing effect prediction model is constructed. The polishing effect prediction model takes the workpiece basic attribute data and initial process parameters as input and outputs the predicted polishing effect data. The basic attribute data of the workpiece to be polished is input into the polishing effect prediction model, and the process parameters are iteratively adjusted in combination with the real-time polishing equipment operation data to generate a dynamic optimization parameter sequence. Based on the aforementioned dynamic optimization parameter sequence, the equipment operation adaptation requirements, process execution sequence, and quality control nodes are integrated to output a complete optimization scheme for the polishing process.

[0006] According to a second aspect of this application, a polishing process optimization system incorporating big data analysis is provided. The polishing process optimization system incorporating big data analysis includes a processor and a readable storage medium storing a program that, when executed by the processor, implements the aforementioned polishing process optimization method incorporating big data analysis.

[0007] Based on any of the above aspects, by integrating workpiece basic attribute data, polishing equipment operation data, process parameter execution data, and polishing effect feedback data, a multi-source data linkage set for the polishing process is constructed. Based on this multi-source data linkage set, the dynamic coupling relationship between different process parameters is mined, forming a process parameter coupling rule set. This reveals the collaborative logic and adaptation constraints between parameters. The polishing effect prediction model constructed based on the process parameter coupling rule set and historical polishing effect feedback data can accurately predict the polishing effect using workpiece basic attribute data and initial process parameters as input. By inputting the basic attribute data of the workpiece to be polished into the prediction model and iteratively adjusting the process parameters in conjunction with real-time polishing equipment operation data, a dynamic optimization parameter sequence is generated. This achieves real-time and dynamic optimization of process parameters, better adapting to changes in different workpieces and actual production conditions. The final output is a complete optimized polishing process scheme that integrates equipment operation adaptation requirements, process execution sequence, and quality control nodes, ensuring the operability and effectiveness of the optimization scheme and comprehensively improving the quality and efficiency of the polishing process. Attached Figure Description

[0008] Figure 1 A schematic flowchart of the polishing process optimization method combining big data analysis provided in the embodiments of this application is shown. Figure 2 A schematic diagram of the component structure of the polishing process optimization system combining big data analysis provided in an embodiment of this application is shown. Detailed Implementation

[0009] Figure 1 A schematic flowchart of the polishing process optimization method combining big data analysis provided in an embodiment of this application is shown, and the detailed steps are described below.

[0010] Step S110: Combine the workpiece basic attribute data, polishing equipment operation data, process parameter execution data and polishing effect feedback data to establish a multi-source data linkage set for the polishing process. The multi-source data linkage set for the polishing process includes real-time synchronization links and historical association indexes for each data type.

[0011] In this embodiment, the polishing process of stainless steel faucets in the hardware and sanitary ware industry is used as an example to construct a multi-source data linkage set for the polishing process. The basic attribute data of the workpiece includes the faucet's material composition information (such as stainless steel type, alloy composition ratio, etc.), shape and structure information (such as main body dimensions, bending angle, surface curvature, etc.), and initial surface state information (such as initial roughness, surface defect distribution, etc.). This data is in the form of a structured data table, with each record corresponding to the basic attributes of a faucet to be polished. The polishing equipment operation data comes from the polishing machine's sensor acquisition system, including equipment operating status information (such as motor speed, equipment vibration frequency, temperature, etc.), actual parameter execution value information (such as actual applied polishing pressure, polishing wheel speed, etc.), and equipment load information (such as current, power, etc.). The data is stored in time series format, with a sampling frequency of 10 times per second. The process parameter execution data covers various set parameters during the polishing process, such as polishing pressure, polishing speed, polishing time, and polishing fluid type, and is recorded in the form of a parameter configuration table, with each process stage corresponding to a set of parameters. The polishing effect feedback data comes from the quality inspection process, including multiple quality indicators such as surface roughness, gloss, and flatness. The data format is structured data corresponding to the test report.

[0012] When establishing a real-time synchronization link, an industrial Ethernet network is used to connect the workpiece basic attribute database, equipment sensor system, process parameter control system, and quality inspection data acquisition system. A data gateway facilitates data protocol conversion between different data sources, ensuring that all data types are transmitted synchronously according to a unified timestamp. For example, when a faucet workpiece enters the polishing station, its basic attribute data is read via an RFID tag, triggering the data synchronization process. At this point, equipment operation data and process parameter execution data are bound to the workpiece ID, forming a real-time data stream. The establishment of a historical association index is achieved by constructing a relational database, using the workpiece ID as the primary key, and storing the workpiece's basic attribute data, equipment operation data at each stage, process parameter execution data, and final polishing effect feedback data in association. Simultaneously, a timestamp and process stage identifier are added to each data entry to quickly locate relevant data at specific times and process stages during subsequent queries and analysis.

[0013] Step S120: Based on the multi-source data linkage set of the polishing process, the dynamic coupling relationship between different process parameters is mined to form a process parameter coupling rule set, which includes parameter change coordination logic and parameter adaptation constraints.

[0014] In this embodiment, based on the multi-source data linkage set for the polishing process constructed above, the dynamic coupling relationship between different process parameters is explored using the polishing process of stainless steel faucets as an example. First, a large amount of historical process parameter execution data is extracted from the multi-source data linkage set. This data covers the polishing process parameters of different batches and models of faucets.

[0015] Step S121: Extract all historical process parameter execution data from the multi-source data linkage set of the polishing process, divide the parameter categories according to the polishing process stage, and each parameter category corresponds to a core execution link in the polishing process.

[0016] In the polishing process of stainless steel faucets, the polishing procedure typically includes three stages: rough polishing, intermediate polishing, and fine polishing. After extracting all historical process parameter execution data from a multi-source data aggregation set for the polishing process, the parameters are categorized according to these three stages. The parameters for the rough polishing stage mainly include rough polishing pressure, rough polishing wheel speed, and rough polishing time; the parameters for the intermediate polishing stage include intermediate polishing pressure, intermediate polishing wheel speed, intermediate polishing time, and intermediate polishing time; and the parameters for the fine polishing stage include fine polishing pressure, fine polishing wheel speed, fine polishing time, and polishing fluid flow rate. Each parameter category corresponds to a core execution step in the polishing process. For example, the rough polishing parameter category corresponds to the initial grinding of the faucet surface, the intermediate polishing parameter category corresponds to the surface imperfection repair step, and the fine polishing parameter category corresponds to the surface brightening step.

[0017] Step S122: Extract the specific process parameter items under each parameter category, and record the complete change trajectory of each process parameter item under different polishing scenarios. The change trajectory includes the parameter's initial value, intermediate adjustment value, and final stable value.

[0018] Taking the rough polishing parameter category as an example, the specific process parameters under it include rough polishing pressure and rough polishing wheel speed. For different polishing scenarios, such as stainless steel faucets with different material thicknesses or different initial surface roughness, the complete change trajectory of these parameters is recorded. For example, for faucets with thicker material, the initial value of the rough polishing pressure may be set to a higher value. During the polishing process, based on real-time feedback on the surface grinding, it may be adjusted multiple times until a stable value is reached. Similarly, the rough polishing wheel speed will have an initial value, intermediate adjustment values ​​due to load changes, and a final stable value. These change trajectories are recorded as a curve with time on the horizontal axis and parameter values ​​on the vertical axis, while also labeling the time point and process state corresponding to each value.

[0019] Step S123: Associate the change trajectories of process parameter items of different parameter categories under the same polishing scenario, establish a parameter change time axis, and make the change nodes of all parameter items synchronously aligned in the time dimension.

[0020] Select a specific polishing scenario, such as polishing a batch of stainless steel faucets of a particular model. Correlate the changes in process parameters across three categories: rough polishing, medium polishing, and fine polishing. Establish a timeline for parameter changes with the polishing start time as the origin. Mark the change nodes for rough polishing pressure and rough polishing wheel speed in the rough polishing category, the change nodes for medium polishing pressure and medium polishing wheel speed in the medium polishing category, and the change nodes for each parameter in the fine polishing category on this timeline. By using the scale on the timeline, ensure that the change nodes for all parameter items are synchronized across the time dimension, thus allowing observation of how parameters from different categories change at the same point in time or within a time interval.

[0021] Step S124: Analyze the relationship between changes in different process parameters within a single parameter category, track the impact path of adjusting the value of one process parameter on other process parameters in the same category, and record the time delay and magnitude of the impact.

[0022] Taking the fine polishing parameter category as an example, this category includes two process parameters: fine polishing pressure and polishing slurry flow rate. When the fine polishing pressure value is adjusted, its impact on the polishing slurry flow rate is traced. For example, when the fine polishing pressure increases, the polishing slurry flow rate may need to be increased accordingly to ensure polishing effect and avoid excessive friction and heat generation. By analyzing the relationship between the two in historical data, the time delay of the impact is determined, i.e., how long after the fine polishing pressure is adjusted will the polishing slurry flow rate begin to adjust, as well as the correlation between the magnitude of the changes, such as the range of increases in polishing slurry flow rate for each increase in fine polishing pressure. The above-mentioned impact path, time delay, and correlation of the magnitude of change are recorded in the form of text descriptions and data tables.

[0023] Step S125: Analyze the cross-influence of process parameter items between different parameter categories, extract cross-category parameter adjustment association cases where changes in process parameter items in one parameter category trigger adjustments in process parameter items in other parameter categories, and generate association rules describing the influence relationship between parameter categories based on the cross-category parameter adjustment association cases.

[0024] Step S1251: Select complete polishing process cases from the multi-source data linkage set of the polishing process, which contain records of changes in process parameter items of multiple parameter categories. The complete polishing process cases contain clear time nodes of parameter changes and corresponding adjustment records.

[0025] From the multi-source data linkage set of polishing process, we selected complete stainless steel faucet polishing process cases that include records of changes in process parameter items for coarse polishing, medium polishing, and fine polishing. These cases need to have clear time nodes for parameter changes, such as the coarse polishing pressure being adjusted from value A to value B at a certain time, as well as corresponding adjustment records, such as the reason for the adjustment and the operator.

[0026] Step S1252: Classify and label the process parameters in each complete polishing process case according to parameter category, and indicate the parameter category to which each process parameter belongs and its functional position in process execution.

[0027] For each selected complete polishing process case, the process parameters are categorized and labeled according to rough polishing, medium polishing, and fine polishing parameters. For example, rough polishing pressure and rough polishing wheel speed are labeled as rough polishing parameters, with the function of rough polishing pressure in the process being to provide grinding pressure, and the function of rough polishing wheel speed being to ensure grinding efficiency. Similarly, the parameters under the medium polishing and fine polishing parameter categories are labeled in a similar way.

[0028] Step S1253: Construct a parameter change tracking map, with time as the horizontal axis and parameter category as the vertical axis, marking the numerical changes of each process parameter item at different time nodes, and intuitively presenting the parameter change trajectory.

[0029] A parameter change tracking graph is constructed with time (in minutes) on the horizontal axis and the parameters for coarse, medium, and fine polishing on the vertical axis. In the graph, the numerical change of each process parameter is marked at the corresponding time point. For example, if the coarse polishing pressure under the coarse polishing parameter category changes from its initial value to a new value at the 5th minute, this change point is marked at the corresponding position on the graph, and the graph is connected to form the change curve for that parameter, thus visually presenting the change trajectory of each parameter.

[0030] Step S1254: Identify the first change node of a process parameter item in a parameter category from the parameter change tracking map, and use the first change node as the starting point for cross-influence analysis.

[0031] In the constructed parameter change tracking map, carefully observe the changes in process parameters under each parameter category. For example, it was found that the rotational speed of the rough polishing wheel in the rough polishing parameter category changed for the first time at the 3rd minute, and this time point was used as the starting point for cross-influence analysis.

[0032] Step S1255: Track the changes in the values ​​of process parameters in other parameter categories after the starting point, and record the category, adjustment time, and adjustment range of the first process parameter item whose value is adjusted.

[0033] Starting from the third minute after the initial change in the coarse polishing wheel speed, track the changes in the process parameters within the intermediate and fine polishing parameter categories. Assuming that the intermediate polishing pressure is adjusted at the fourth minute, record the intermediate polishing parameter category to which this process parameter belongs, the adjustment time of the fourth minute, and the adjustment range, such as how much it has been adjusted from the initial value of the intermediate polishing pressure.

[0034] Step S1256: Analyze the correlation between changes in initial parameters and subsequent parameter adjustments, extract cross-category parameter adjustment correlation cases directly caused by changes in initial parameters, and record the parameter category combination, initial parameter change characteristics, subsequent parameter adjustment characteristics, and corresponding polishing scene information in the cross-category parameter adjustment correlation cases.

[0035] Analyze the correlation between the initial change in the coarse polishing wheel speed and the subsequent adjustment of the intermediate polishing pressure. By reviewing relevant process records and equipment operating data, determine whether the adjustment of the intermediate polishing pressure is directly caused by the change in the coarse polishing wheel speed. If a direct correlation is found, extract this as a cross-category parameter adjustment correlation case. Record the parameter category combination in this case as coarse polishing parameter category and intermediate polishing parameter category, the initial parameter change characteristic as the coarse polishing wheel speed is adjusted from one value to another at the 3rd minute, the subsequent parameter adjustment characteristic as the intermediate polishing pressure is adjusted accordingly at the 4th minute, and the corresponding polishing scenario information, such as the polishing of a specific model of stainless steel faucet.

[0036] Step S1257: Classify the cross-category parameter adjustment association cases, and divide the case types according to the combination of the starting parameter category and the affected parameter category. Each case type contains multiple similar cross-category parameter adjustment association cases.

[0037] The extracted cross-category parameter adjustment correlation cases are categorized. For example, cases where the initial parameter category is coarse-throw parameter category and the affected parameter category is medium-throw parameter category are classified into one case type; cases where the initial parameter category is medium-throw parameter category and the affected parameter category is fine-throw parameter category are classified into another case type. Each case type contains multiple cross-category parameter adjustment correlation cases with similar parameter category combinations.

[0038] Step S1258: In each case type, statistically analyze the correspondence between the initial parameter change range and the subsequent parameter adjustment range, and calculate and record the numerical ratio and change pattern between the initial parameter change range and the subsequent parameter adjustment range.

[0039] Taking a case type where the initial parameter category is coarse-throwing parameter category and the affected parameter category is intermediate-throwing parameter category as an example, we statistically analyze the correspondence between the change range of coarse-throwing parameters and the adjustment range of intermediate-throwing parameters in all cases under this type. We calculate the numerical ratio between the change range of the initial parameters and the adjustment range of subsequent parameters, such as the numerical ratio by which the intermediate-throwing pressure increases for every certain increase in the coarse-throwing wheel speed, and record the variation pattern of the above ratio in different cases, such as whether the ratio is fixed or fluctuates within a certain range.

[0040] Step S1259: Analyze the differences in the influence of cross-category parameters under different polishing scenarios in the same case type, extract the moderating effect of scenario features on the degree of influence, and label the influence weight of scenario factors.

[0041] Within the same case type, different polishing scenarios may lead to differences in the impact of cross-category parameters. For example, for stainless steel faucets with different material thicknesses, the degree to which changes in the coarse polishing wheel speed affects the adjustment of the intermediate polishing pressure may vary. Analyzing these differences, we extract the moderating effect of scenario features such as material thickness and initial surface roughness on the degree of influence, and label the influence weight of each scenario factor based on historical data, such as the influence weight of material thickness being 0.6 and the influence weight of initial surface roughness being 0.4.

[0042] Step S12510: Analyze the common data patterns of cross-category parameter influence in each case type, including influence delay time, adjustment magnitude ratio, and parameter change direction relationship, and generate preliminary influence association rules based on this.

[0043] We analyze common data patterns in the cross-category parameter impact across each case type. For example, in a case type where the initial parameter category is coarse-throwing parameters and the affected parameter category is intermediate-throwing parameters, the impact delay time is typically around one minute, the adjustment magnitude ratio is within a fixed range, and the parameter changes in the same direction—that is, as the coarse-throwing wheel speed increases, the intermediate-throwing pressure also increases. Based on these common data patterns, we generate preliminary impact association rules. For instance, when the coarse-throwing wheel speed in the coarse-throwing parameter category increases by a certain magnitude, the intermediate-throwing pressure in the intermediate-throwing parameter category will increase in the same direction according to a certain proportional relationship after one minute.

[0044] Step S12511: Integrate the preliminary influence association rules of all case types, supplement the special adaptation items of influence under special scenarios, and generate a cross-category parameter influence association rule set. The cross-category parameter influence association rule set is used to describe the cross-influence relationship of process parameter items between different parameter categories.

[0045] The initial influence association rules for all case types are integrated into a unified rule set. Simultaneously, considering special scenarios such as abnormal polishing fluid temperature and equipment aging, corresponding special adaptation items are added. For example, when the polishing fluid temperature is too high, the adjustment ratio of cross-category parameter influence needs to be adjusted. Finally, a cross-category parameter influence association rule set is generated, which details the cross-influence relationships of process parameter items between different parameter categories.

[0046] Step S126: For polishing pressure-related parameters and polishing speed-related parameters, collect and compare the collaborative process cases of their changes under different workpiece materials, extract the collaborative logic of parameter changes, and record the order of changes and numerical ratios of the two.

[0047] This study collects data on the coordinated process of polishing pressure-related parameters (such as rough polishing pressure, intermediate polishing pressure, and fine polishing pressure) and polishing speed-related parameters (such as rough polishing wheel speed, intermediate polishing wheel speed, and fine polishing wheel speed) under different workpiece materials during the polishing process of stainless steel faucets. For example, for faucets made of 304 stainless steel and 201 stainless steel, data on the changes in polishing pressure and polishing speed during the polishing process are collected. By comparing the above cases, the coordinated logic of parameter changes is extracted. It is found that when polishing 304 stainless steel faucets, the polishing speed is usually adjusted first, and then the polishing pressure is adjusted according to the polishing effect. The numerical ratio between the two is that for every certain increase in polishing speed, the polishing pressure increases by a corresponding value. However, when polishing 201 stainless steel faucets, the polishing pressure may be adjusted first, and then the polishing speed is adjusted, and the numerical ratio is also different. The order of changes and the numerical ratio of these two parameters under these different materials are recorded.

[0048] Step S127: For the parameters related to polishing time and the parameters related to polishing fluid type, compare the adaptation process cases under different workpiece shapes, determine the parameter adaptation constraints, and record the applicable scope and limitation boundaries of the combined use of the parameters related to polishing time and the parameters related to polishing fluid type.

[0049] Polishing time-related parameters include rough polishing time, intermediate polishing time, and fine polishing time. Polishing fluid type-related parameters include acidic polishing fluid, neutral polishing fluid, and alkaline polishing fluid. Compare suitable process cases for different workpiece shapes, such as the combined use of polishing time and polishing fluid type for faucets with large surface curvature and faucets with relatively flat surfaces. Determine the constraints of parameter adaptation. For example, for faucets with large surface curvature, the rough polishing time should not be too long when using acidic polishing fluid, otherwise it will lead to excessive local corrosion; while for faucets with flat surfaces, the fine polishing time can be appropriately extended when using neutral polishing fluid. Record the applicable range of these polishing time-related parameters and polishing fluid type-related parameters, such as the suitability of acidic polishing fluid for faucets with large surface curvature within a certain range of rough polishing time, and the limiting boundaries, such as the damage to the workpiece if the rough polishing time of acidic polishing fluid exceeds a certain value.

[0050] Step S128: Integrate the parameter association logic, association rules, parameter change coordination logic, and parameter adaptation constraints within a single category to form preliminary parameter coupling rules.

[0051] The parameter association logic within a single parameter category obtained in step S124, the association rules obtained in step S125, the parameter change coordination logic refined in step S126, and the parameter adaptation constraints determined in step S127 are integrated. For example, the association logic between fine polishing pressure and polishing fluid flow rate within the fine polishing parameter category, the association rules between coarse and medium polishing parameter categories, the coordination logic between polishing pressure and polishing speed changes, and the adaptation constraints between polishing time and polishing fluid type are integrated to form preliminary parameter coupling rules.

[0052] Step S129: Traverse the historical polishing process cases in the multi-source data linkage set of the polishing process, and verify the applicability of the preliminary parameter coupling rules with actual parameter changes and effect feedback data.

[0053] The process iterates through a large number of historical stainless steel faucet polishing process cases in a multi-source data linkage set. The actual parameter changes in each case are compared with the preliminary parameter coupling rules. Simultaneously, the applicability of the preliminary parameter coupling rules is verified by incorporating the polishing effect feedback data for that case. For example, it checks whether the changes in polishing pressure and polishing speed in a particular case conform to the refined collaborative logic. If they do, and the polishing effect is good, then the rule is applicable; if they do not, or the polishing effect is poor, then the rule may have a problem.

[0054] Step S1210: Optimize the preliminary parameter coupling rules based on the verification results, supplement special adaptation items for parameter coupling under special working conditions, and correct rule content that does not match the actual polishing process case.

[0055] Based on the verification results, the initial parameter coupling rules were optimized. Rules that were found to be inconsistent with actual polishing process cases during verification were revised, such as adjusting the numerical proportions or order of parameter changes. Simultaneously, special adaptation items for parameter coupling under specific operating conditions were added, such as rules on how to adjust the polishing speed and time when a minor equipment malfunction causes unstable polishing pressure.

[0056] Step S1211: Classify and organize the optimized parameter coupling rules according to parameter category and coupling type to form a structured process parameter coupling rule set. Add scene identifiers to the process parameter coupling rule set to indicate the applicable scope of workpiece material, shape and equipment type corresponding to each parameter coupling rule. The process parameter coupling rule set includes parameter change coordination logic and parameter adaptation constraints.

[0057] The optimized parameter coupling rules are categorized and organized according to parameter types such as coarse polishing, medium polishing, and fine polishing, as well as coupling types such as parameter change coordination and parameter adaptation constraints, forming a structured set of process parameter coupling rules. A scenario identifier is added to each parameter coupling rule, clearly indicating the applicable scope for the rule's corresponding workpiece material (e.g., 304 stainless steel, 201 stainless steel), workpiece shape (e.g., faucet with large surface curvature, faucet with a flat surface), and equipment type (e.g., a specific model of polishing machine). The final set of process parameter coupling rules contains rich parameter change coordination logic and parameter adaptation constraints, which can be used to guide subsequent polishing process optimization.

[0058] Step S130: Based on the process parameter coupling rule set and historical polishing effect feedback data, construct a polishing effect prediction model. The polishing effect prediction model takes the workpiece basic attribute data and initial process parameters as input and outputs predicted polishing effect data.

[0059] In this embodiment, a polishing effect prediction model is constructed by combining the process parameter coupling rule set of the stainless steel faucet polishing process with historical polishing effect feedback data. The historical polishing effect feedback data includes a large amount of data on the surface roughness, gloss, and flatness of stainless steel faucets under different combinations of process parameters.

[0060] Step S131: Extract historical polishing effect feedback data from the multi-source data linkage set of the polishing process, and split the data content according to the polishing quality evaluation dimension. Each polishing quality evaluation dimension corresponds to a core polishing effect indicator.

[0061] Historical polishing effect feedback data was extracted from a multi-source data aggregation dataset of polishing processes. This data covers the polishing effects of multiple batches of stainless steel faucets. The data content was broken down according to polishing quality evaluation dimensions, such as surface roughness, gloss, and flatness. Each polishing quality evaluation dimension corresponds to a core polishing effect indicator, such as the surface roughness dimension corresponding to the surface roughness value indicator, the gloss dimension corresponding to the gloss value indicator, and the flatness dimension corresponding to the flatness error indicator.

[0062] Step S132: Associate the historical polishing effect feedback data with the corresponding process parameter execution data and workpiece basic attribute data to form an associated data group for each polishing process case.

[0063] The extracted historical polishing effect feedback data is correlated with the process parameter execution data that produced that effect (such as polishing pressure, speed, and time at each stage) and the corresponding workpiece basic attribute data (such as material, shape, and initial surface condition). For example, in the polishing effect feedback data of a batch of stainless steel faucets, the surface roughness is Ra1.2. This data is correlated with the process parameter execution data such as rough polishing pressure, rough polishing wheel speed, and intermediate polishing time during the polishing of that batch of faucets, as well as the workpiece basic attribute data such as the stainless steel type and main body dimensions of that batch of faucets, forming a correlated data group corresponding to a polishing process case.

[0064] Step S133: Extract the parameter coupling rules corresponding to each associated data group from the set of process parameter coupling rules, and label the collaborative logic and constraints followed by parameter changes in the associated data group.

[0065] For each associated data group, extract the corresponding parameter coupling rules from the process parameter coupling rule set. For example, if an associated data group involves changes in coarse polishing pressure and coarse polishing wheel speed, find the corresponding parameter change coordination logic from the rule set, such as the rule on how the coarse polishing pressure should be adjusted when the coarse polishing wheel speed increases, and mark the coordination logic followed by the parameter changes in this associated data group, as well as the related constraints, such as the adjustment range of the coarse polishing pressure.

[0066] Step S134: Construct the basic model structure, which includes an attribute feature input layer, a parameter coupling rule embedding layer, an effect index prediction layer, and a result output layer. Information flows between the layers through a data transmission channel.

[0067] The basic structure for constructing a polishing effect prediction model comprises four main layers. The attribute feature input layer receives and initially processes the workpiece's basic attribute data; the parameter coupling rule embedding layer fuses the process parameter execution data with the corresponding parameter coupling rules; the effect index prediction layer predicts various polishing effect indicators based on the fused data; and the result output layer outputs the prediction results in a preset format. Information flows orderly between layers through data transmission channels, ensuring accurate and efficient data transfer and processing between them.

[0068] Step S135: Convert the material features and shape features in the basic attribute data of the workpiece into feature vectors that the model can recognize, and use them as the standard input format for the attribute feature input layer.

[0069] For material features in the workpiece's basic attribute data, such as stainless steel grade, they are converted into uniquely coded vectors, with different stainless steel grades corresponding to different coding combinations. Shape features, such as main dimensions and bending angles, are standardized and arranged in a certain order to form vectors. The converted material feature vectors and shape feature vectors are concatenated to form a feature vector that the model can recognize, which serves as the standard input format for the attribute feature input layer. For example, the material feature vector is [0,1,0] (representing 304 stainless steel), and the shape feature vector is [standardized main dimension value, standardized bending angle value,...]. After concatenation, a comprehensive attribute feature vector is formed.

[0070] Step S136: Convert the process parameter execution data into parameter vectors according to parameter categories, fuse them with the corresponding parameter coupling rule codes to form a parameter-rule fusion vector, and input it into the parameter coupling rule embedding layer.

[0071] The process parameter execution data is converted into parameter vectors according to parameter categories such as coarse polishing, intermediate polishing, and fine polishing. For example, parameters such as coarse polishing pressure, coarse polishing wheel speed, and coarse polishing time in the coarse polishing parameter category are standardized to form a coarse polishing parameter vector; the intermediate and fine polishing parameter categories are processed similarly. Simultaneously, the corresponding parameter coupling rules are encoded, such as converting parameter change coordination logic and constraints into specific encoded vectors. Then, the parameter vectors and rule encoded vectors are fused, for example, by element-wise addition, to form a parameter-rule fusion vector, which is then input into the parameter coupling rule embedding layer.

[0072] Step S137: In the parameter coupling rule embedding layer, the attribute feature vector and the parameter-rule fusion vector are processed by the feature interaction algorithm to explore the nonlinear correlation between the two and generate the correlation feature matrix.

[0073] Step S1371: Perform dimensional expansion processing on the attribute feature vector of the input parameter coupling rule embedding layer to make the dimension of the attribute feature vector consistent with the dimension of the parameter-rule fusion vector. Use a feature mapping algorithm to map the dimension-expanded attribute feature vector to a high-dimensional feature space and transform it into a high-dimensional attribute feature representation.

[0074] The attribute feature vector of the input parameter-coupled rule embedding layer has a dimension of D1, and the parameter-rule fusion vector has a dimension of D2. If D1 is not equal to D2, the attribute feature vector is expanded in dimension, for example, by adding a zero vector to expand its dimension to D2. Then, a feature mapping algorithm, such as multinomial mapping, is used to map the dimension-expanded attribute feature vector to a high-dimensional feature space, transforming it into a high-dimensional attribute feature representation. For example, the original two-dimensional attribute feature vector is mapped to a ten-dimensional high-dimensional space, and the value of each dimension is obtained through different combinations of the original vectors.

[0075] Step S1372: For the parameter-rule fusion vector, the feature components corresponding to the parameter coupling rule encoding are weighted by weighting operation to increase the weight of the rule features in feature interaction.

[0076] The parameter-rule fusion vector is formed by fusing a parameter vector and a rule encoding vector, where the feature components corresponding to the rule encoding vector have higher importance in feature interactions. Through weighting operations, the feature components corresponding to the rule encoding vector are assigned higher weight values, while the feature components corresponding to the parameter vector are assigned relatively lower weight values. For example, the weight of the feature component corresponding to the rule encoding vector is set to 1.5, and the weight of the feature component corresponding to the parameter vector is set to 1.0. Weighting is achieved by multiplying each component by its corresponding weight value.

[0077] Step S1373: Construct a dual-path feature interaction channel. One feature interaction channel is used to handle the direct interaction between attribute features and parameter features, and the other feature interaction channel is used to handle the indirect interaction between attribute features and rule features.

[0078] A dual-path feature interaction channel is constructed. In the direct interaction channel, the high-dimensional attribute feature representation interacts with the parametric feature part of the parameter-rule fusion vector; in the indirect interaction channel, the high-dimensional attribute feature representation interacts with the rule feature part of the parameter-rule fusion vector. This allows for the capture of the interaction relationships between attribute features and parametric features, and between attribute features and rule features, respectively.

[0079] Step S1374: In the direct interaction feature interaction channel, element-wise multiplication is used to achieve element-wise interaction between the high-dimensional attribute feature representation and the parametric feature part in the weighted parameter-rule fusion vector, generating a direct interaction feature vector; in the indirect interaction feature interaction channel, matrix multiplication is used to achieve matrix-level interaction between the high-dimensional attribute feature representation and the regular feature part in the weighted parameter-rule fusion vector, generating an indirect interaction feature matrix.

[0080] In the direct interaction channel, element-wise multiplication is performed between the high-dimensional attribute feature representation and the parametric feature portion of the weighted parameter-rule fusion vector. This involves multiplying corresponding elements to generate the direct interaction feature vector. For example, multiplying the i-th element of the high-dimensional attribute feature representation with the i-th element of the parametric feature portion yields the i-th element of the direct interaction feature vector. In the indirect interaction channel, matrix multiplication is performed between the high-dimensional attribute feature representation (as a matrix) and the rule feature portion of the weighted parameter-rule fusion vector (as a matrix) to generate the indirect interaction feature matrix. Matrix multiplication follows standard matrix multiplication rules: the rows of the previous matrix are multiplied element-wise by the corresponding columns of the next matrix, and then the results are summed.

[0081] Step S1375: Use a feature weighting algorithm to assign weights to the direct interaction feature vector and the indirect interaction feature matrix. The weight values ​​are set based on the degree of influence of attribute features, parameter features and rule features on the polishing effect in historical data.

[0082] Based on the influence of attribute features, parameter features, and rule features on the polishing effect in historical data, weight values ​​are assigned to the direct interaction feature vector and the indirect interaction feature matrix. For example, analysis of historical data reveals that rule features have a significant impact on the polishing effect; therefore, a higher weight value, such as 0.6, is assigned to the indirect interaction feature matrix, and a weight value of 0.4 is assigned to the direct interaction feature vector. A feature weighting algorithm is used, such as multiplying each element of the direct interaction feature vector by 0.4 and each element of the indirect interaction feature matrix by 0.6, to achieve the weight allocation.

[0083] Step S1376: Concatenate the weighted direct interaction feature vector and the indirect interaction feature matrix by dimension to form a preliminary interaction feature set, which contains feature information under different interaction modes.

[0084] The weighted direct interaction feature vector (a one-dimensional vector) and the indirect interaction feature matrix (a two-dimensional matrix) are concatenated dimensionally. For example, the direct interaction feature vector is added as a row below the indirect interaction feature matrix to form a preliminary interaction feature set. In this way, the preliminary interaction feature set contains information on both the direct interaction between attribute features and parameter features, and the indirect interaction between attribute features and rule features.

[0085] Step S1377: Select features from the preliminary interactive feature set that are more than a preset threshold in relation to the polishing effect prediction target using a feature selection algorithm, and use them as key features; group the key features according to their correlation with different polishing effect indicators to form grouped key features.

[0086] Feature selection algorithms, such as mutual information-based feature selection methods, are employed to calculate the correlation between each feature in the initial interactive feature set and the polishing effect prediction target (such as surface roughness, gloss, etc.). A preset threshold, such as 0.3, is set, and features with a correlation higher than this threshold are selected as key features. Then, these key features are grouped according to their correlation with different polishing effect indicators. For example, key features highly correlated with surface roughness are grouped into one group, and key features highly correlated with gloss are grouped into another group, forming the grouped key features.

[0087] Step S1378: The grouped key features are transformed into a structured association feature matrix using a matrix reconstruction algorithm, generating an association feature matrix containing all key interaction features. The association feature matrix is ​​used to reflect the nonlinear association between attribute features, parameter features and rule features. The row dimension of the association feature matrix corresponds to the feature category, the column dimension corresponds to the sample identifier, and the matrix elements are feature values.

[0088] The key features after grouping are processed using a matrix reconstruction algorithm. The key features of different groups are arranged in a certain order, with rows corresponding to feature categories (e.g., attribute feature groups, parameter feature groups, rule feature groups), and columns corresponding to sample identifiers. Each sample corresponds to one column, and the matrix elements are the corresponding feature values. Through this method, the grouped key features are transformed into a structured correlation feature matrix, which reflects the non-linear relationships between attribute features, parameter features, and rule features.

[0089] Step S138: Input the associated feature matrix into the effect index prediction layer. The effect index prediction layer independently predicts each polishing effect index through a multi-dimensional regression algorithm, generates a single index prediction value, integrates all single index prediction values ​​to form complete predicted polishing effect data, and outputs it through the result output layer in a preset format.

[0090] The associated feature matrix is ​​input into the performance indicator prediction layer, which independently predicts each polishing performance indicator (such as surface roughness, gloss, and smoothness) using a multi-dimensional regression algorithm. For example, for the surface roughness indicator, a linear regression algorithm is used, taking relevant features from the associated feature matrix as input and calculating the predicted value of surface roughness using the trained model parameters; for the gloss indicator, a decision tree regression algorithm may be used for prediction. All individual indicator predictions are then integrated to form complete predicted polishing performance data, and the prediction results are output through the results output layer in a preset format, such as JSON.

[0091] Step S139: Collect multiple sets of related data as training samples, input the training samples into the basic structure of the model for iterative training, and update the algorithm parameters and feature interaction weights of each layer through iterative training.

[0092] For example, step S1391: extract a sufficient number of associated data groups from the multi-source data linkage set of the polishing process, clean the extracted associated data groups, and obtain effective associated data groups.

[0093] A large number of related data sets are extracted from the multi-source data set of the polishing process. The number should be sufficient for model training, such as several thousand sets. The extracted related data sets are then cleaned to remove missing, outlier, and duplicate values. Missing values ​​can be filled using mean imputation or interpolation; outliers are identified and removed or corrected using methods such as box plots; and duplicate records are directly deleted. After cleaning, valid related data sets are obtained.

[0094] Step S1392: Divide the effective associated data group into a training dataset, a validation dataset, and a test dataset according to a preset ratio. The training dataset is used for model parameter training, the validation dataset is used for parameter adjustment during the training process, and the test dataset is used for the final performance evaluation of the model.

[0095] The effective associated data sets are divided into training, validation, and test datasets in a preset ratio of 7:2:1. The training dataset contains most of the associated data sets and is used to train the algorithm parameters and feature interaction weights of each layer of the model; the validation dataset is used to evaluate the model performance during training and adjust the model parameters based on the evaluation results; the test dataset is used for the final performance evaluation after the model training is completed to test the model's generalization ability.

[0096] Step S1393: Set the objective function for model training. The objective function aims to minimize the error between the model's predicted polishing effect data and the actual polishing effect data, thereby quantifying the training effect of the model.

[0097] The objective function is defined as the mean squared error function, which is calculated as the average of the sum of the squares of the differences between the predicted value of each indicator in the model's predicted polishing effect data and the actual value of the corresponding indicator in the actual polishing effect data. The goal is to minimize the value of this mean squared error function through training, thereby quantifying the training effect of the model.

[0098] Step S1394: Initialize the algorithm parameters and feature interaction weights of each layer of the model infrastructure. The initial values ​​of the parameters are randomly initialized, and the initial values ​​of the weights are initially allocated based on the importance of the features.

[0099] The algorithm parameters of the attribute feature input layer, parameter coupling rule embedding layer, effect index prediction layer, and result output layer in the model's basic structure are randomly initialized, with initial values ​​randomly selected within a small range, such as between -0.01 and 0.01. The initial values ​​of feature interaction weights are initially assigned based on feature importance; features that have a greater impact on polishing effect in historical data are assigned higher initial weight values, while those that have less impact are assigned lower initial weight values.

[0100] Step S1395: Input the training dataset into the model infrastructure in batches. After each batch of data is input, calculate the model's predicted output, i.e., the predicted polishing effect data, through the forward propagation algorithm.

[0101] The training dataset is divided into multiple batches, each containing a certain number of associated data groups, such as 50. The data is input into the model infrastructure batch by batch. Through the forward propagation algorithm, the data passes sequentially from the attribute feature input layer through the parameter coupling rule embedding layer and the effect index prediction layer, finally obtaining the predicted polishing effect data in the result output layer.

[0102] Step S1396: Compare the predicted polishing effect data with the actual polishing effect data in the training dataset, calculate the error value between the two through the objective function, and quantify the current prediction accuracy of the model.

[0103] The predicted polishing effect data output by the model is compared with the corresponding actual polishing effect data in the training dataset. The predicted and actual values ​​of each indicator are substituted into the objective function (mean squared error function) to calculate the error value between the two. The smaller the error value, the higher the current prediction accuracy of the model.

[0104] Step S1397: Use the backpropagation algorithm to propagate the error value from the result output layer to the attribute feature input layer, and calculate the contribution of each layer's algorithm parameters and feature interaction weights to the error layer by layer.

[0105] A backpropagation algorithm is employed, starting from the output layer and propagating the calculated error value back to the attribute feature input layer in the reverse direction of data propagation. During this propagation, the contribution of each layer's algorithm parameters and feature interaction weights to the error is calculated layer by layer, i.e., the partial derivatives of the error value with respect to each parameter and weight are calculated.

[0106] Step S1398: Based on the contribution of each layer's algorithm parameters and feature interaction weights to the error, update the algorithm in the direction of reducing the error and control the step size with the learning rate parameter.

[0107] Based on the contribution (partial derivative) of each layer's algorithm parameters and feature interaction weights to the error, the parameters and weights are updated according to the direction of gradient descent. The learning rate parameter controls the update step size; a small learning rate, such as 0.001, is set to ensure the stability of parameter updates. Through iterative calculation, the parameters and weights are continuously adjusted to gradually reduce the error value of the objective function.

[0108] Step S1399: After completing the training of a batch of data, input the validation dataset into the model infrastructure, calculate the prediction error of the model infrastructure on the validation dataset, and evaluate the performance of the model after parameter adjustment.

[0109] After training a batch of data, a validation dataset is input into the model's infrastructure. The forward propagation algorithm is used to obtain predicted polishing effect data, which is then compared with the actual polishing effect data in the validation dataset to calculate the prediction error. Based on this prediction error, the performance of the model after parameter adjustment is evaluated to determine if the model has improved.

[0110] Step S13910: If the verification error meets the preset error threshold, continue training on the next batch of data; if the verification error does not meet the threshold, adjust the learning rate parameter and the objective function coefficient, and re-optimize the parameters.

[0111] A preset error threshold, such as 0.05, is set. If the validation error is less than or equal to this threshold, the model performance meets the requirements, and training continues on the next batch of data. If the validation error is greater than this threshold, the learning rate parameter needs to be adjusted, such as reducing the learning rate to 0.5 times the original value, or adjusting the objective function coefficients, such as increasing the coefficient of the regularization term. Then, the parameters are optimized again until the validation error meets the threshold.

[0112] Step S13911: Repeat the above batch training, parameter adjustment and validation evaluation steps until the entire training dataset has been trained and the performance of the model infrastructure on the validation dataset reaches the preset standard.

[0113] Following the steps described above, train the model on all batches of training data. After each batch is completed, adjust the parameters and perform validation evaluation. Repeat this process until the entire training dataset is trained and the model's performance on the validation dataset meets the preset standard, such as the validation error stabilizing below a threshold and no longer significantly decreasing.

[0114] Step S13912: Input the test dataset into the trained model infrastructure, and calculate the prediction accuracy and mean error of the model infrastructure on the test dataset.

[0115] Input the test dataset into the trained model infrastructure to obtain predicted polishing effect data. Compare the predicted data with the actual data in the test dataset to calculate the prediction accuracy, which is the percentage of samples where the deviation between the predicted and actual values ​​is within the allowable range; at the same time, calculate the mean error, such as the mean absolute error.

[0116] Step S13913: Based on the evaluation results of the test dataset, perform final calibration on the algorithm parameters and feature interaction weights of each layer of the model's basic structure, complete the iterative training of the model, and obtain the parameters and weights of the trained polishing effect prediction model.

[0117] Based on the evaluation results of the test dataset, such as prediction accuracy and mean error, the algorithm parameters and feature interaction weights of each layer of the model's basic structure are finally calibrated. For example, if the parameters of a certain layer contribute significantly to the prediction error, the parameters of that layer are fine-tuned. After calibration, the model iterative training ends, and the parameters and weights of the trained polishing effect prediction model are obtained.

[0118] Step S1310: Set up a cross-validation process, divide the training samples into multiple subsets, select the hierarchical structure and algorithm type of the model basic structure based on the training and validation results of different subsets, combine the trained model basic structure, the updated algorithm parameters and the selected structure configuration for validation, and construct the polishing effect prediction model. The polishing effect prediction model takes the workpiece basic attribute data and initial process parameters as input and outputs the predicted polishing effect data.

[0119] A cross-validation process, such as five-fold cross-validation, is implemented. The training samples are divided into five subsets, with four subsets selected sequentially as the training set and one subset as the validation set, for five training and validation iterations. Based on the results of these five training and validation iterations with different subsets, the hierarchical structure of the model's basic architecture (e.g., the number of neurons in each layer, the number of layers) and the algorithm type (e.g., the choice of regression algorithm) are selected. For example, by comparing the validation errors of models under different hierarchical structures, the hierarchical structure with the smallest error is selected. The trained model's basic architecture, the updated algorithm parameters, and the selected validation structure are combined to finally construct a polishing effect prediction model. This polishing effect prediction model takes the workpiece's basic attribute data and initial process parameters as input, and after processing through each layer, outputs predicted polishing effect data.

[0120] Step S140: Input the basic attribute data of the workpiece to be polished into the polishing effect prediction model, and iteratively adjust the process parameters in combination with the real-time polishing equipment operation data to generate a dynamic optimization parameter sequence.

[0121] In this embodiment, taking a new stainless steel faucet workpiece as an example, its basic attribute data is input into a trained polishing effect prediction model, and the process parameters are iteratively adjusted by combining real-time polishing equipment operation data to generate a dynamic optimization parameter sequence.

[0122] Step S141: Collect basic attribute data of the workpiece to be polished. The basic attribute data includes workpiece material composition information, shape and structure information, and initial surface state information, and is consistent with the dimensions of the workpiece basic attribute data in the multi-source data linkage set of the polishing process.

[0123] Collect basic attribute data of the stainless steel faucet to be polished, including material composition information (e.g., stainless steel type 304, alloy composition with 18% chromium and 8% nickel), shape and structure information (main body dimensions are 20 cm long, 10 cm wide, and 15 cm high, bending angle is 90 degrees, surface curvature radius is 5 cm, etc.), and initial surface condition information (initial roughness is Ra3.2, and there are 2 scratches on the surface, etc.). The dimensions of the above data are consistent with the dimensions of the workpiece basic attribute data in the multi-source data linkage set of the polishing process so that the subsequent model can process it correctly.

[0124] Step S142: Convert the basic attribute data of the workpiece to be polished into a standard feature vector, and perform data preprocessing according to the input format requirements of the polishing effect prediction model.

[0125] The collected basic attribute data of the workpiece to be polished is converted into standard feature vectors. For the stainless steel type in the material composition information, unique thermal encoding is used to convert it into a vector; numerical data such as alloy composition percentage are standardized and mapped to the range of 0-1. Data such as dimensions, angles, and curvature in the shape and structure information are also standardized and formed into vectors. Roughness values ​​in the initial surface condition information are standardized and added to the vector; discrete data such as the number of surface scratches are also converted into corresponding codes. These vectors are concatenated in a specific order to form a standard feature vector, and then checked and adjusted according to the input format requirements of the polishing effect prediction model, such as data type and dimensions, to ensure compliance with the model's input requirements.

[0126] Step S143: Set the initial set of process parameters for the workpiece to be polished. The initial set of process parameters includes polishing pressure-related parameters, polishing speed-related parameters, polishing time-related parameters, and polishing fluid type-related parameters. The initial values ​​of the parameters are set based on conventional polishing process experience.

[0127] Based on experience with conventional polishing processes for stainless steel faucets, an initial set of process parameters was set. Polishing pressure parameters included: coarse polishing pressure set to 2.5 N, medium polishing pressure set to 1.8 N, and fine polishing pressure set to 1.2 N; polishing speed parameters included: coarse polishing wheel speed set to 1500 rpm, medium polishing wheel speed set to 2000 rpm, and fine polishing wheel speed set to 2500 rpm; polishing time parameters included: coarse polishing time set to 5 minutes, medium polishing time set to 3 minutes, and fine polishing time set to 2 minutes; and polishing fluid type parameters were set to neutral polishing fluid.

[0128] Step S144: Input the standard feature vector and the initial process parameter set into the polishing effect prediction model. The polishing effect prediction model outputs the corresponding predicted polishing effect data, which includes the predicted values ​​of each polishing effect index.

[0129] The transformed standard feature vector and the set initial process parameters are input into the polishing effect prediction model. After processing through various internal layers, the model outputs corresponding predicted polishing effect data. This predicted polishing effect data includes predicted values ​​for various polishing effect indicators, such as predicted surface roughness of Ra0.8, predicted gloss of 85 gloss units, and predicted flatness error of 0.05 mm.

[0130] Step S145: Collect real-time operating data after the polishing equipment is started. The real-time operating data includes equipment operating status information, actual parameter execution value information, and equipment load information, and is consistent with the dimensions of the equipment operating data in the multi-source data linkage set of the polishing process.

[0131] After the polishing equipment is started, it collects operating data in real time through sensors on the equipment. The equipment operating status information includes motor speed (actual measured value), equipment vibration frequency (e.g., 20 Hz), equipment temperature (e.g., 45 degrees Celsius), etc.; the actual executed parameter values ​​include actual coarse polishing pressure (e.g., 2.3 N) and actual coarse polishing wheel speed (e.g., 1480 rpm), etc.; the equipment load information includes current (e.g., 5 Amps) and power (e.g., 1.5 kW), etc. The dimensions of the above real-time operating data are consistent with the dimensions of the equipment operating data in the multi-source data linkage set of the polishing process.

[0132] Step S146: Compare the actual execution values ​​of parameters in the real-time running data with the parameter values ​​in the initial process parameter set, extract the deviation information between the two, match the deviation information based on the process parameter coupling rule set, identify the deviation type and infer its influence trend on the polishing effect.

[0133] The actual values ​​of parameters in the real-time operating data are compared with the parameter values ​​in the initial process parameter set. For example, the actual value of the rough polishing pressure is 2.3 N, the initial setting is 2.5 N, and the deviation is -0.2 N; the actual value of the rough polishing wheel speed is 1480 rpm, the initial setting is 1500 rpm, and the deviation is -20 rpm. After extracting these deviation information, matching is performed based on the process parameter coupling rule set. According to the relevant rules in the rule set regarding the deviation of polishing pressure and polishing speed, the deviation type is identified, such as insufficient pressure and low speed. The trend of its impact on the polishing effect is inferred, such as insufficient pressure and low speed may lead to insufficient polishing removal, resulting in a surface roughness higher than the predicted value and a gloss lower than the predicted value.

[0134] Step S147: Based on the deviation type, influence trend, predicted polishing effect data and the process parameter coupling rule set, determine the process parameter item to be adjusted, and calculate the adjustment direction based on the parameter change coordination logic and adaptation constraints.

[0135] Based on the identified deviation types (insufficient pressure, low rotation speed), influence trends (increased surface roughness, decreased gloss), predicted polishing effect data (current predicted surface roughness Ra0.8), and the process parameter coupling rule set, the process parameters to be adjusted are determined to be rough polishing pressure and rough polishing wheel rotation speed. According to the parameter change coordination logic in the rule set, such as polishing pressure and polishing speed should be adjusted in the same direction to ensure polishing effect, and adapting to constraints, such as the adjustment range of rough polishing pressure being 2.0-3.0 N and the adjustment range of rough polishing wheel rotation speed being 1400-1600 rpm, the adjustment direction is calculated to be increasing rough polishing pressure and increasing rough polishing wheel rotation speed.

[0136] Step S148: Calculate the adjustment range for each process parameter to be adjusted. The adjustment range is set based on the degree of deviation, the gap between the predicted polishing effect and the expected effect, and the constraint requirements in the parameter coupling rules.

[0137] Step S1481: Construct an adjustment range calculation process, which includes a deviation impact quantification step, an effect gap quantification step, a rule constraint quantification step, and a comprehensive calculation step. Each step works together to complete the adjustment range calculation.

[0138] A calculation process for the adjustment range is constructed, which sequentially includes steps for quantifying the impact of deviations, quantifying the effect gap, quantifying rule constraints, and conducting a comprehensive calculation. These four steps work together to calculate the adjustment range of the process parameter to be adjusted.

[0139] Step S1482: In the deviation impact quantification step, the degree of deviation between the actual executed value of the parameter in the real-time running data and the initial process parameter is converted into a quantitative index, which is used to reflect the magnitude of the impact of the deviation on the process execution.

[0140] In the deviation impact quantification step, for a rough polishing pressure deviation of -0.2 N, the relationship between pressure deviation and process impact in historical data is used to convert it into a quantitative index. For example, for every 0.1 N pressure deviation, the quantitative index increases by 0.2, so the quantitative index corresponding to a deviation of -0.2 N is 0.4. This quantitative index reflects the magnitude of the deviation's impact on process execution.

[0141] Step S1483: In the effect gap quantification step, calculate the difference between the predicted value of each indicator in the predicted polishing effect data and the preset expected effect indicator value, and convert the difference into a standardized effect gap quantification value. The size of the quantification value corresponds to the degree of significance of the effect gap.

[0142] Preset expected performance indicators, such as an expected surface roughness of Ra 0.6 and an expected gloss of 90 gloss units. Calculate the difference between the predicted and expected values ​​of each indicator in the predicted polishing effect data. For example, the surface roughness difference is 0.8 - 0.6 = 0.2Ra, and the gloss difference is 85 - 90 = -5 gloss units. Standardize these differences; for example, convert the surface roughness difference of 0.2Ra to a quantitative value of 0.3, and the gloss difference of -5 gloss units to a quantitative value of 0.2. The larger the quantitative value, the more significant the difference in performance.

[0143] Step S1484: In the rule constraint quantification step, the adaptation constraint conditions related to the process parameter item to be adjusted are extracted from the process parameter coupling rule set, and the constraint conditions are converted into quantified constraint values. The quantified constraint values ​​are used to reflect the allowable range of parameter adjustment.

[0144] The adaptation constraints related to the coarse polishing pressure and the speed of the coarse polishing wheel are extracted from the process parameter coupling rule set. For example, the adjustment range of the coarse polishing pressure is 2.0-3.0 N, the current actual pressure is 2.3 N, which is 0.7 N away from the upper limit and 0.3 N away from the lower limit. The above constraints are converted into quantitative constraint values. For example, the pressure adjustment space of 0.7 N is converted into a quantitative constraint value of 0.7, which reflects the allowable range of parameter adjustment.

[0145] Step S1485: Standardize the deviation impact quantification index, effect gap quantification value, and rule constraint quantification value based on a unified evaluation scale to obtain standardized deviation impact value, effect gap value, and rule constraint value.

[0146] The quantified indicators of the impact of bias (e.g., 0.4), the quantified values ​​of the effect gap (e.g., the average of 0.3 and 0.2, 0.25), and the quantified values ​​of the rule constraints (e.g., 0.7) are standardized based on a uniform evaluation scale (e.g., 0-1). For example, the quantified indicator of the impact of bias of 0.4, which is within the original range (0-1), remains 0.4 after standardization; the quantified value of the effect gap of 0.25, which is within the original range (0-1), becomes 0.25 after standardization; and the quantified value of the rule constraints of 0.7, which is within the original range (0-1), becomes 0.7 after standardization.

[0147] Step S1486: Set the weight coefficients for the deviation impact value, effect gap value, and rule constraint value. The weight coefficients are set based on the importance of each factor to the parameter adjustment and historical adjustment experience.

[0148] Based on the importance of each factor to parameter adjustment and historical adjustment experience, the weighting coefficients for the deviation impact value, effect gap value, and rule constraint value are set at 0.4, 0.4, and 0.2 respectively. These weighting coefficients reflect the importance of each factor in the adjustment range calculation.

[0149] Step S1487: Multiply the standardized deviation impact value, effect gap value, and rule constraint value by their corresponding weight coefficients to obtain the weighted evaluation results of each factor. In the comprehensive calculation step, sum the weighted evaluation results of each factor to obtain the comprehensive evaluation value of parameter adjustment. The comprehensive evaluation value is used to reflect the necessity and adjustment space of parameter adjustment.

[0150] Multiplying the standardized bias impact value of 0.4 by the weighting coefficient of 0.4 yields 0.16; multiplying the effect gap value of 0.25 by the weighting coefficient of 0.4 yields 0.1; multiplying the rule constraint value of 0.7 by the weighting coefficient of 0.2 yields 0.14. In the comprehensive calculation step, these three weighted evaluation results are summed: 0.16 + 0.1 + 0.14 = 0.4, resulting in a comprehensive evaluation value of 0.4 for parameter adjustment. This comprehensive evaluation value reflects the necessity of parameter adjustment and the adjustment space.

[0151] Step S1488: Construct an adjustment range mapping function. The adjustment range mapping function takes the comprehensive evaluation value as input and the parameter adjustment range as output. The functional relationship is set based on the correspondence between the comprehensive evaluation value and the actual adjustment range in a large number of historical parameter adjustment process cases.

[0152] An adjustment range mapping function is constructed. By analyzing the correspondence between the comprehensive evaluation value and the actual adjustment range in a large number of historical parameter adjustment process cases, the functional relationship is determined. For example, when the comprehensive evaluation value is 0.4, the corresponding adjustment range of the coarse polishing pressure is 0.3 Newtons, and the adjustment range of the coarse polishing wheel speed is 30 revolutions per minute.

[0153] Step S1489: Input the comprehensive evaluation value of parameter adjustment into the adjustment range mapping function to obtain the preliminary adjustment range value, wherein the preliminary adjustment range value is the reference range of parameter adjustment.

[0154] The comprehensive evaluation value of 0.4 for parameter adjustment is input into the adjustment range mapping function to obtain the initial adjustment range value of 0.3 N for coarse polishing pressure and 30 rpm for coarse polishing wheel speed. These initial adjustment range values ​​are used as the reference range for parameter adjustment.

[0155] Step S14810: Combine the current value range of the process parameter to be adjusted, perform boundary verification on the preliminary adjustment range value, so that the adjusted parameter value is between the maximum and minimum range allowed by the process.

[0156] The current value of the rough polishing pressure, a process parameter to be adjusted, is 2.3 N. The maximum allowable range is 3.0 N, and the minimum is 2.0 N. The initial adjustment increment is 0.3 N, resulting in a pressure of 2.3 + 0.3 = 2.6 N, which is within the 2.0-3.0 N range. Boundary verification is passed. The current value of the rough polishing wheel speed is 1480 rpm, with an allowable range of 1400-1600 rpm. The initial adjustment increment is 30 rpm, resulting in a speed of 1480 + 30 = 1510 rpm, also within the allowable range. Boundary verification is passed.

[0157] Step S14811: Correct the initial adjustment range value based on the boundary verification results. If the initial adjustment range value causes the parameter to exceed the boundary, the adjustment range is corrected to the difference or sum of the boundary value and the current parameter value. The corrected adjustment range value is then fine-tuned by referring to the historical parameter adjustment range data of similar workpieces to make the adjustment range more in line with the actual process execution requirements. The final adjustment range of each process parameter item to be adjusted is then determined.

[0158] Since the initial adjustment range value boundary verification passed in this embodiment, no correction is required. Referring to the historical parameter adjustment range data of similar stainless steel faucets, it was found that a coarse agitation pressure adjustment of 0.3 N and a rotation speed adjustment of 30 rpm are appropriate under similar circumstances. Therefore, no fine-tuning is performed on the initial adjustment range value, and the final adjustment range of the coarse agitation pressure is determined to be 0.3 N, and the final adjustment range of the coarse agitation wheel rotation speed is determined to be 30 rpm.

[0159] Step S149: Update the initial process parameter set according to the adjustment direction and adjustment range to generate the adjusted process parameter set, and input the adjusted process parameter set into the polishing effect prediction model to obtain new predicted polishing effect data.

[0160] The initial process parameter set is updated according to the adjustment direction (increase) and adjustment range (coarse polishing pressure 0.3 N, coarse polishing wheel speed 30 rpm). The updated coarse polishing pressure is 2.5 + 0.3 = 2.8 N, and the coarse polishing wheel speed is 1500 + 30 = 1530 rpm, while other parameters remain unchanged, generating an adjusted process parameter set. This adjusted process parameter set, along with the standard feature vector of the workpiece to be polished, is input into the polishing effect prediction model to obtain new predicted polishing effect data, such as a predicted surface roughness of Ra 0.7 and a gloss of 88 gloss units.

[0161] Step S1410: Continuously collect real-time polishing equipment operation data, repeat the above steps of parameter comparison, deviation analysis, adjustment direction determination, adjustment range calculation and parameter update, and form an iterative adjustment cycle of process parameters.

[0162] Continuously collect real-time data on the polishing equipment operating under the adjusted process parameters, such as an actual rough polishing pressure of 2.7 N and an actual rotation speed of 1520 rpm. Repeat the steps of parameter comparison (comparing with the adjusted set of process parameters and calculating the deviation), deviation analysis (identifying the type of deviation and its impact trend), determining the adjustment direction, calculating the adjustment range, and updating parameters. For example, if new deviations still exist, continue adjusting the parameters to form an iterative adjustment cycle for process parameters.

[0163] Step S1411: Set the iteration termination condition. When the predicted polishing effect data meets the preset effect threshold and the parameter deviation in the real-time equipment operation data is within the preset range, stop the iteration adjustment, integrate all the adjusted process parameter sets during the iteration process, sort them according to the adjustment time sequence and polishing process stage, and form a dynamic optimization parameter sequence that includes parameter change trajectory, adjustment trigger conditions and corresponding equipment operation status.

[0164] Set iteration termination conditions, such as predicting surface roughness below Ra0.6, gloss above 90 gloss units, and parameter deviations in real-time equipment operating data (e.g., pressure deviation within ±0.1 N, speed deviation within ±20 rpm) within preset ranges. Stop iterative adjustment when these conditions are met. Integrate all adjusted process parameters from the iteration process, sorting them according to adjustment time sequence and the roughing, intermediate, and fine polishing stages to form a dynamic optimization parameter sequence. This dynamic optimization parameter sequence includes the parameter change trajectory at each stage (e.g., roughing pressure adjusted from 2.5 N to 2.8 N and then to 2.7 N), the triggering conditions for each adjustment (e.g., deviation type, influence trend), and the corresponding equipment operating status (e.g., vibration frequency, temperature).

[0165] Step S150: Based on the dynamic optimization parameter sequence, integrate equipment operation adaptation requirements, process execution timing and quality control nodes to output a complete optimization scheme for the polishing process.

[0166] In this embodiment, based on the above-generated dynamic optimization parameter sequence for polishing stainless steel faucets, the equipment operation adaptation requirements, process execution sequence, and quality control nodes are integrated to output a complete optimization scheme for the polishing process.

[0167] Step S151: Extract parameter configuration information for each process stage from the dynamic optimization parameter sequence, divide the process execution units according to the order of polishing processes, and each process execution unit corresponds to a set of parameter configurations and a process execution stage.

[0168] Parameter configuration information for each process stage—rough polishing, intermediate polishing, and fine polishing—is extracted from the dynamic optimization parameter sequence. The parameter configuration for the rough polishing stage includes rough polishing pressure, rough polishing wheel speed, and rough polishing time; the intermediate polishing stage includes intermediate polishing pressure, intermediate polishing wheel speed, and intermediate polishing time; and the fine polishing stage includes fine polishing pressure, fine polishing wheel speed, fine polishing time, and polishing fluid flow rate. Based on the sequence of polishing processes, the above parameter configuration information is divided into rough polishing process execution units, intermediate polishing process execution units, and fine polishing process execution units. Each process execution unit corresponds to a set of parameter configurations and one process execution stage.

[0169] Step S152: Analyze the parameter configuration information of each process execution unit, combine it with the equipment operation data in the multi-source data linkage set of the polishing process, retrieve the equipment operation records that match each parameter configuration from the historical equipment operation data, and extract the standard operation action sequence and parameter setting value.

[0170] Analyze the parameter configuration information of the rough polishing process execution unit, such as rough polishing pressure of 2.8 N and rotation speed of 1530 rpm. Combine this with the equipment operation data in the multi-source data linkage set of the polishing process, and retrieve equipment operation records that match the parameter configuration from historical equipment operation data. For example, find operation records of past rough polishing using similar parameter configurations, and extract the standard operation sequence, such as the order and duration of actions like starting the polishing wheel, adjusting the distance between the polishing wheel and the workpiece, and starting the feed, as well as parameter settings, such as the initial rotation speed setting of the polishing wheel and the gradual increase method of pressure.

[0171] Step S153: Based on the standard operation sequence and parameter settings, generate a set of equipment control instructions for each process execution unit. The set of equipment control instructions includes the specified equipment operation mode, the standard operation sequence, and the parameter calibration target value.

[0172] Based on the extracted standard operating sequence and parameter settings, a set of equipment control instructions is generated for each process execution unit. For the rough polishing process execution unit, the equipment control instruction set includes specifying that the equipment operates in rough polishing mode. The standard operating sequence is: start the polishing wheel to 1530 rpm, adjust the distance between the polishing wheel and the workpiece to 5 mm, apply 2.8 N of pressure, and maintain this state for 5 minutes, etc. The parameter calibration target values ​​are that the rough polishing pressure is stabilized at 2.8 N ± 0.1 N, and the rotation speed is stabilized at 1530 rpm ± 20 rpm.

[0173] Step S154: Based on the time nodes for parameter adjustment in the dynamic optimization parameter sequence and the order of process execution units, plan the process execution sequence and mark the start time, duration, and connection time of each process execution unit.

[0174] The dynamic optimization parameter sequence includes the time nodes for adjusting each parameter. Combined with the order of process execution units (rough polishing → intermediate polishing → fine polishing), the process execution sequence is planned. The start time of the rough polishing process execution unit is marked as 0 minutes, and the duration is 5 minutes; the start time of the intermediate polishing process execution unit is 5 minutes, and the duration is 3 minutes; the start time of the fine polishing process execution unit is 8 minutes, and the duration is 2 minutes. The connection time between each process execution unit is also marked; for example, the connection time from the end of rough polishing to the start of intermediate polishing is 0 minutes, meaning that intermediate polishing begins immediately after rough polishing.

[0175] Step S155: Extract historical polishing quality control data from the multi-source data linkage set of the polishing process, combine it with the parameter change nodes in the dynamic optimization parameter sequence, determine the quality control node of each process execution unit, and mark the control object and control method.

[0176] Historical polishing quality control data is extracted from the multi-source data set of the polishing process. For example, surface roughness is typically checked after the rough polishing stage, and surface defects are checked after the intermediate polishing stage. Combining parameter change nodes in the dynamic optimization parameter sequence, such as rough polishing pressure adjustment nodes and intermediate polishing speed adjustment nodes, quality control nodes for each process execution unit are determined. For instance, the quality control node for the rough polishing process execution unit is at the end of rough polishing, with surface roughness being the control object, and the control method being surface roughness inspection using a roughness meter; the quality control node for the intermediate polishing process execution unit is at the end of intermediate polishing, with surface defects being the control object, and the control method being visual inspection.

[0177] Step S156: Configure the corresponding quality inspection program and inspection trigger conditions for each quality control node.

[0178] Configure a quality inspection program for the quality control node of the rough polishing process execution unit. For example, use a roughness tester to uniformly select five inspection points on the workpiece surface for measurement, and take the average value as the inspection result. The inspection is triggered when the rough polishing process execution unit's duration ends, i.e., at the 5th minute. Configure a vision inspection program for the quality control node of the intermediate polishing process execution unit. Use an industrial camera to capture images of the workpiece surface, and perform image processing analysis to determine the number and size of defects. The inspection is triggered when the intermediate polishing process execution unit's duration ends, i.e., at the 8th minute.

[0179] Step S157: Integrate the parameter configuration information, equipment operation adaptation requirements, process execution sequence and quality control nodes of each process execution unit to form a process execution unit plan. Each process execution unit plan corresponds to a complete process execution link.

[0180] The parameter configuration information (pressure, speed, time, etc.), equipment operation adaptation requirements (equipment operation mode, operation sequence, etc.), process execution sequence (start time 0 minutes, duration 5 minutes), and quality control nodes (roughness detection at the end) of the rough polishing process execution unit are integrated to form a rough polishing process execution unit scheme. Similarly, the relevant information of the intermediate polishing and fine polishing process execution units are integrated to form their respective process execution unit schemes, with each scheme corresponding to a complete process execution stage.

[0181] Step S158: Connect all process execution unit schemes in sequence according to the process execution time to construct the main process execution flow, and mark the connection logic and data transmission requirements between process execution units.

[0182] The process execution units for rough polishing, intermediate polishing, and fine polishing are connected in sequence according to the process execution time (0-5 minutes for rough polishing, 5-8 minutes for intermediate polishing, and 8-10 minutes for fine polishing) to construct the main process execution flow. The connection logic between the process execution units is marked, such as after rough polishing, the workpiece is automatically transferred to the intermediate polishing station, and the intermediate polishing equipment starts the intermediate polishing process execution unit upon receiving the workpiece. Data transmission requirements include transmitting the quality inspection results from the rough polishing stage to the intermediate polishing stage as a reference for fine-tuning the intermediate polishing parameters; and transmitting the quality inspection results from the intermediate polishing stage to the fine polishing stage, also as a basis for fine polishing parameter adjustment.

[0183] Step S159: Extract equipment maintenance data corresponding to the dynamic optimization parameter sequence from the multi-source data linkage set of the polishing process, and formulate equipment pre-maintenance suggestions and process maintenance plans in combination with the main process execution flow.

[0184] Equipment maintenance data corresponding to equipment operating parameters (such as speed and pressure) in the dynamic optimization parameter sequence are extracted from the multi-source data linkage set of the polishing process. This includes data such as the wear condition of the polishing wheel at that speed and the maintenance cycle of the bearings. Based on the main process execution flow, pre-process maintenance recommendations are formulated, such as checking the wear degree of the polishing wheel before the process begins and replacing it if the wear exceeds the threshold; checking the equipment lubrication system and adding lubricating oil, etc. The process maintenance plan includes a simple cleaning and tightening check of the equipment every 2 hours during the rough polishing stage; and real-time monitoring of equipment temperature and vibration throughout the entire process, with shutdown and maintenance initiated if any abnormalities occur.

[0185] Step S1510: Based on the main process flow, equipment pre-maintenance recommendations, and process maintenance plan, estimate the human, time, and material resources required for the polishing process and generate a resource requirement list.

[0186] Based on the main process flow, the entire polishing process takes 10 minutes per piece. Pre-process maintenance requires one maintenance personnel to spend 30 minutes on inspection and preparation. In-process maintenance requires one maintenance personnel to conduct inspections every 2 hours, for 10 minutes each time. Material resources include polishing wheels (each usable for 50 workpieces), polishing fluid (1 liter consumed per 10 workpieces), etc. The estimated human resource requirement is 2 personnel (1 operator, 1 maintenance personnel), time resources are 10 minutes of processing time per workpiece plus pre-process preparation and maintenance time, and material resources are the specific quantities of polishing wheels, polishing fluid, etc., forming a resource requirements list.

[0187] Step S1511: Construct an emergency response process during process execution, defining the triggering conditions, execution steps, and execution entities for possible equipment failures, parameter deviations, and quality anomalies.

[0188] Step S1511-1: Extract fault case data from the multi-source data linkage set of the polishing process in the history of polishing, and count the possible equipment fault types, parameter deviation types and quality anomaly types to form a fault anomaly type list.

[0189] Fault case data from historical polishing processes are extracted from the multi-source data linkage set of polishing process, and possible equipment fault types are statistically identified, such as polishing wheel jamming, motor overheating, sensor failure, etc.; parameter deviation types, such as sudden increase or decrease in pressure, excessive speed fluctuation, etc.; and quality abnormality types, such as increased surface scratches, substandard gloss, and out-of-tolerance flatness, etc., forming a fault abnormality type list.

[0190] Step S1511-2: For each type of equipment failure, analyze the common causes, characteristics, and impact on process execution, and extract effective failure handling steps based on historical experience.

[0191] For the equipment failure type of polishing wheel jamming, common causes include foreign object entrapment and improper installation of the polishing wheel. Symptoms include the polishing wheel suddenly stopping rotation and the equipment emitting abnormal noises. The impact on process execution is that the workpiece currently being polished may be scrapped, and subsequent processes cannot proceed normally. Based on historical experience, effective troubleshooting steps are summarized as follows: immediately stop the machine, disconnect the power supply, check for and remove any foreign objects around the polishing wheel, check the polishing wheel's installation for security, reinstall or replace the polishing wheel, and resume production after a trial run confirms the fault has been resolved.

[0192] Step S1511-3: For each type of parameter deviation, analyze the triggering conditions, changing trends, and degree of influence on the polishing effect of the deviation, and formulate deviation correction measures in combination with the process parameter coupling rule set.

[0193] For the parameter deviation type of sudden pressure increase, the triggering conditions may include hydraulic system failure, sensor false alarms, etc.; the trend is that the pressure value rises rapidly in a short period of time; the impact on the polishing effect is that it may lead to over-polishing of the workpiece surface, resulting in dents or burns. Based on the process parameter coupling rule set, deviation correction measures are formulated: immediately stop the pressure increase, check for leaks or blockages in the hydraulic system, calibrate the pressure sensor, and if immediate repair is not possible, switch to the backup hydraulic system and adjust other relevant parameters (such as reducing the rotational speed) according to the rule set to mitigate the impact of excessive pressure.

[0194] Step S1511-4: For each type of quality anomaly, analyze the key nodes where the anomaly occurs, the propagation path, and the impact on the final polishing quality, and formulate anomaly containment measures and quality remediation plans.

[0195] For the quality anomaly of increased surface scratches, analysis suggests the key point of occurrence may be in the rough polishing stage; the propagation path is that scratches may not be completely eliminated in the intermediate and fine polishing stages, leading to final product defects; the impact on final polishing quality is that the product appearance does not meet requirements, necessitating rework or scrapping. Anomaly containment measures are implemented: immediately stop the rough polishing process, inspect the polishing wheel surface for hard objects or damage, replace the polishing wheel, and mark the workpieces with scratches. Quality remediation plan: add a fine grinding process to the marked workpieces to attempt to repair the scratches; if repair is impossible, scrap the workpiece.

[0196] Step S1511-5: Set the response priority for each type of fault / abnormality. The response priority is set based on the degree of impact of the fault / abnormality on the process schedule, quality, and cost.

[0197] Set response priorities for different fault / abnormality types, categorized into high, medium, and low levels. For example, a stuck polishing wheel that renders the equipment inoperable and significantly impacts process progress has a high response priority; slight pressure fluctuations have a minor impact on polishing results and have a low response priority; and increased surface scratches affect product quality and have a medium response priority.

[0198] Step S1511-6: Delineate the responsible departments and personnel for emergency response, and mark the scope of fault and anomaly handling and the responsibilities and authority of each responsible entity.

[0199] Emergency response responsibilities are divided between the Production Department and the Equipment Department. The Production Department is responsible for handling quality anomalies, such as increased surface scratches or substandard gloss, and the responsible personnel are production team leaders and quality inspectors. The Equipment Department is responsible for handling equipment malfunctions and parameter deviations, and the responsible personnel are equipment engineers and maintenance workers. The scope of responsibility and authority of each responsible entity are clearly defined; for example, equipment engineers are responsible for diagnosing complex equipment malfunctions and developing repair plans, while maintenance workers are responsible for carrying out repair operations.

[0200] Step S1511-7: Define the emergency response process, which includes an automated logical chain of fault and anomaly triggering, measure matching and execution, and feedback of processing results.

[0201] Define the emergency response process: When an abnormality is detected (such as abnormal pressure detected by sensor data or surface scratches found by quality inspectors), the system automatically triggers the emergency response process; the system matches the corresponding response measures from the emergency response rule base according to the type of abnormality; the responsible personnel execute the measures after receiving the notification; after the processing is completed, the processing results are fed back to the system, and the system records the processing process and results, forming an automated logical chain of closed-loop management.

[0202] Step S1511-8: Set emergency handling time thresholds and mark the maximum allowable time for handling each type of fault / abnormality.

[0203] Emergency handling time thresholds are set for different types of malfunctions. For example, the maximum allowable handling time for a stuck polishing wheel is 30 minutes to reduce the impact on production schedule; the maximum allowable handling time for increased surface scratches is 20 minutes to promptly curb the spread of the abnormality; and the maximum allowable handling time for slight pressure fluctuations is 15 minutes.

[0204] Step S1511-9: Combine the fault handling steps, deviation correction measures, anomaly containment measures, response priority, execution entity definition, emergency handling process and time threshold to generate a preliminary emergency handling rule base; update the preliminary emergency handling rule base based on historical processing data, add special scenario rules, correct execution logic, and generate an emergency handling rule base containing handling solutions for all fault and anomaly types.

[0205] The aforementioned fault handling steps, deviation correction measures, anomaly containment measures, response priorities, execution entity definitions, emergency handling procedures, and time thresholds are combined to generate a preliminary emergency handling rule base. Then, based on historical processing data, the preliminary rule base is updated, such as adding special fault handling rules for specific equipment models and correcting the execution logic order of certain measures, ultimately generating an emergency handling rule base containing handling solutions for all types of faults and anomalies.

[0206] Step S1512: Integrate the main process execution flow, equipment operation adaptation requirements, quality control nodes, equipment maintenance suggestions, resource requirement list and emergency handling procedures to form a complete optimization plan for the polishing process, and output it in a standardized format.

[0207] This comprehensive optimization plan for the polishing process integrates the main process flow, equipment operation adaptation requirements for each process unit, quality control nodes, equipment pre-maintenance recommendations and process maintenance plans, resource requirements (human resources, time resources, and material resources), and emergency response procedures. The plan should be output in a standardized format, including sections such as plan name, process stage division, parameter configuration table, operating steps, quality control requirements, resource requirements, and emergency response measures.

[0208] Furthermore, Figure 2 A schematic diagram of the hardware structure of a polishing process optimization system 100 incorporating big data analysis for implementing the methods provided in the embodiments of this application is shown. Figure 2 As shown, the polishing process optimization system 100 combining big data analysis may include at least one processor 102 (the processor 102 may be, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, a transmission device 106 for communication functions, and a controller 108. Those skilled in the art will understand that... Figure 2 The structure shown is for illustrative purposes only and does not limit the structure of the polishing process optimization system 100 incorporating big data analytics. For example, the polishing process optimization system 100 incorporating big data analytics may also include components that are more complex than... Figure 2 The more or fewer components shown, or having the same Figure 2 The different configurations shown.

[0209] The memory 104 can be used to store software programs and modules for application software, such as the program instructions corresponding to the method embodiments described above in this application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the above-described polishing process optimization method combined with big data analysis. The transmission device 106 is used to acquire or send data via a network.

[0210] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

Claims

1. A polishing process optimization method combining big data analysis, characterized in that, The method includes: By combining workpiece basic attribute data, polishing equipment operation data, process parameter execution data, and polishing effect feedback data, a multi-source data linkage set for polishing process is established. The multi-source data linkage set for polishing process includes real-time synchronization links and historical association indexes for each data type. Based on the multi-source data linkage set of the polishing process, the dynamic coupling relationship between different process parameters is mined to form a process parameter coupling rule set, which includes parameter change coordination logic and parameter adaptation constraints. Based on the process parameter coupling rule set and historical polishing effect feedback data, a polishing effect prediction model is constructed. The polishing effect prediction model takes the workpiece basic attribute data and initial process parameters as input and outputs the predicted polishing effect data. The basic attribute data of the workpiece to be polished is input into the polishing effect prediction model, and the process parameters are iteratively adjusted in combination with the real-time polishing equipment operation data to generate a dynamic optimization parameter sequence. Based on the aforementioned dynamic optimization parameter sequence, the equipment operation adaptation requirements, process execution sequence, and quality control nodes are integrated to output a complete optimization scheme for the polishing process.

2. The polishing process optimization method combining big data analysis according to claim 1, characterized in that, The process, based on the multi-source data linkage set of the polishing process, mines the dynamic coupling relationship between different process parameters to form a set of process parameter coupling rules, including: Extract all historical process parameter execution data from the multi-source data linkage set of the polishing process, divide the parameter categories according to the polishing process stage, and each parameter category corresponds to a core execution link in the polishing process. Extract the specific process parameter items under each parameter category, and record the complete change trajectory of each process parameter item under different polishing scenarios. The change trajectory includes the parameter's initial value, intermediate adjustment value, and final stable value. By associating the change trajectories of process parameters of different parameter categories under the same polishing scenario, a parameter change time axis is established so that the change nodes of all parameter items are synchronized and aligned in the time dimension. Analyze the relationship between changes in different process parameters within a single parameter category, track the impact path of adjusting the value of one process parameter on other process parameters in the same category, and record the time delay and magnitude of the impact. The cross-influence of process parameters across different parameter categories is analyzed. Cross-category parameter adjustment correlation cases are extracted, where changes in process parameters in one parameter category trigger adjustments in process parameters in other parameter categories. Based on these cross-category parameter adjustment correlation cases, association rules describing the influence relationship between parameter categories are generated. For polishing pressure-related parameters and polishing speed-related parameters, collaborative process cases of their changes under different workpiece materials are collected and compared. The collaborative logic of parameter changes is extracted, and the order of their changes and their numerical ratios are recorded. For the parameters related to polishing time and polishing fluid type, compare the adapted process cases under different workpiece shapes, determine the parameter adaptation constraints, and record the applicable scope and limiting boundaries of the combined use of polishing time-related parameters and polishing fluid type parameters; Integrate parameter association logic, association rules, parameter change coordination logic, and parameter adaptation constraints within a single category to form preliminary parameter coupling rules; The historical polishing process cases in the multi-source data linkage set of the polishing process are traversed, and the applicability of the preliminary parameter coupling rules is verified by actual parameter changes and effect feedback data. Based on the verification results, the initial parameter coupling rules were optimized, special adaptation items for parameter coupling under special working conditions were added, and rule content that did not match the actual polishing process case was corrected. The optimized parameter coupling rules are categorized and organized according to parameter type and coupling type to form a structured set of process parameter coupling rules. Scenario identifiers are added to the set of process parameter coupling rules to indicate the applicable scope of workpiece material, shape, and equipment type for each parameter coupling rule. The set of process parameter coupling rules includes parameter change coordination logic and parameter adaptation constraints.

3. The polishing process optimization method combining big data analysis according to claim 1, characterized in that, The polishing effect prediction model is constructed based on the process parameter coupling rule set and historical polishing effect feedback data. This model takes workpiece basic attribute data and initial process parameters as input and outputs predicted polishing effect data, including: Historical polishing effect feedback data is extracted from the multi-source data linkage set of the polishing process, and the data content is split according to the polishing quality evaluation dimension. Each polishing quality evaluation dimension corresponds to a core polishing effect indicator. The historical polishing effect feedback data is associated with the corresponding process parameter execution data and workpiece basic attribute data to form a related data group for each polishing process case; Extract the parameter coupling rules corresponding to each associated data group from the set of process parameter coupling rules, and label the collaborative logic and constraints followed by parameter changes in the associated data group; A basic model structure is constructed, which includes an attribute feature input layer, a parameter coupling rule embedding layer, an effect index prediction layer, and a result output layer. Information flows between the layers through a data transmission channel. The material and shape features in the basic attribute data of the workpiece are transformed into feature vectors that the model can recognize, which serve as the standard input format for the attribute feature input layer. The process parameter execution data is converted into parameter vectors according to parameter categories, and then fused with the corresponding parameter coupling rule codes to form a parameter-rule fusion vector, which is then input into the parameter coupling rule embedding layer. In the parameter coupling rule embedding layer, the attribute feature vector and the parameter-rule fusion vector are processed by the feature interaction algorithm to explore the nonlinear correlation between the two and generate the correlation feature matrix. The associated feature matrix is ​​input into the effect index prediction layer. The effect index prediction layer independently predicts each polishing effect index through a multi-dimensional regression algorithm, generates a single index prediction value, integrates all single index prediction values, forms complete predicted polishing effect data, and outputs it through the result output layer in a preset format. Collect multiple sets of related data as training samples, input the training samples into the basic structure of the model for iterative training, and update the algorithm parameters and feature interaction weights of each layer through iterative training; A cross-validation process is set up to divide the training samples into multiple subsets. Based on the training and validation results of different subsets, the hierarchical structure and algorithm type of the model's basic structure are selected. The trained model's basic structure, the updated algorithm parameters, and the selected structure configuration are combined to construct the polishing effect prediction model. The polishing effect prediction model takes the workpiece's basic attribute data and initial process parameters as input and outputs predicted polishing effect data.

4. The polishing process optimization method combining big data analysis according to claim 3, characterized in that, In the parameter coupling rule embedding layer, the attribute feature vector and the parameter-rule fusion vector are processed through a feature interaction algorithm to mine the nonlinear correlation between them and generate a correlation feature matrix, including: The dimension of the attribute feature vector of the input parameter coupling rule embedding layer is expanded to make the dimension of the attribute feature vector consistent with the dimension of the parameter-rule fusion vector. The dimension-expanded attribute feature vector is then mapped to a high-dimensional feature space using a feature mapping algorithm, thus transforming it into a high-dimensional attribute feature representation. For the parameter-rule fusion vector, a weighting operation is used to weight the feature components corresponding to the parameter coupling rule encoding, thereby increasing the weight of the rule features in feature interaction; Construct a dual-path feature interaction channel: one path is used to handle the direct interaction between attribute features and parameter features, and the other path is used to handle the indirect interaction between attribute features and rule features. In the direct interaction feature interaction channel, element-wise multiplication is used to achieve element-wise interaction between the high-dimensional attribute feature representation and the parametric feature part in the weighted parameter-rule fusion vector, generating a direct interaction feature vector; in the indirect interaction feature interaction channel, matrix multiplication is used to achieve matrix-level interaction between the high-dimensional attribute feature representation and the regular feature part in the weighted parameter-rule fusion vector, generating an indirect interaction feature matrix. A feature weighting algorithm is used to assign weights to the direct interaction feature vector and the indirect interaction feature matrix. The weight values ​​are set based on the degree of influence of attribute features, parameter features and rule features on the polishing effect in historical data. The weighted direct interaction feature vector and the indirect interaction feature matrix are concatenated by dimension to form a preliminary interaction feature set, which contains feature information under different interaction modes. The feature selection algorithm selects features from the initial interactive feature set that are more than a preset threshold in relation to the polishing effect prediction target, and uses them as key features. The key features are then grouped according to their correlation with different polishing effect indicators to form grouped key features. A matrix reconstruction algorithm is used to transform the grouped key features into a structured association feature matrix, generating an association feature matrix containing all key interaction features. The association feature matrix is ​​used to reflect the nonlinear relationship between attribute features, parameter features and rule features. The row dimension of the association feature matrix corresponds to the feature category, the column dimension corresponds to the sample identifier, and the matrix elements are feature values.

5. The polishing process optimization method combining big data analysis according to claim 1, characterized in that, The process involves inputting the basic attribute data of the workpiece to be polished into the polishing effect prediction model, iteratively adjusting the process parameters based on real-time polishing equipment operating data, and generating a dynamically optimized parameter sequence, including: Collect basic attribute data of the workpiece to be polished. The basic attribute data includes workpiece material composition information, shape and structure information, and initial surface state information, and is consistent with the dimensions of the workpiece basic attribute data in the multi-source data linkage set of the polishing process. The basic attribute data of the workpiece to be polished is converted into a standard feature vector, and the data is preprocessed according to the input format requirements of the polishing effect prediction model. Set an initial set of process parameters for the workpiece to be polished. The initial set of process parameters includes parameters related to polishing pressure, polishing speed, polishing time, and polishing fluid type. The initial values ​​of the parameters are set based on conventional polishing process experience. The standard feature vector and the initial set of process parameters are input into the polishing effect prediction model, and the polishing effect prediction model outputs the corresponding predicted polishing effect data, which includes the predicted values ​​of each polishing effect index. Real-time operating data of the polishing equipment after startup is collected. The real-time operating data includes equipment operating status information, actual parameter execution value information, and equipment load information, and is consistent with the dimensions of the equipment operating data in the multi-source data linkage set of the polishing process. By comparing the actual execution values ​​of parameters in the real-time running data with the parameter values ​​in the initial process parameter set, the deviation information between the two is extracted. Based on the process parameter coupling rule set, the deviation information is matched to identify the deviation type and infer its influence trend on the polishing effect. Based on the deviation type, influence trend, predicted polishing effect data, and the process parameter coupling rule set, the process parameter items to be adjusted are determined, and the adjustment direction is calculated based on the parameter change coordination logic and adaptation constraints. Calculate the adjustment range for each process parameter to be adjusted. The adjustment range is set based on the degree of deviation, the gap between the predicted polishing effect and the expected effect, and the constraint requirements in the parameter coupling rules. The initial process parameter set is updated according to the adjustment direction and adjustment range to generate the adjusted process parameter set. The adjusted process parameter set is then input into the polishing effect prediction model to obtain new predicted polishing effect data. Continuously collect real-time polishing equipment operation data, repeat the above steps of parameter comparison, deviation analysis, adjustment direction determination, adjustment range calculation and parameter update, and form an iterative adjustment cycle of process parameters; Set an iteration termination condition. When the predicted polishing effect data meets the preset effect threshold and the parameter deviation in the real-time equipment operation data is within the preset range, stop the iteration adjustment, integrate all the adjusted process parameter sets during the iteration process, sort them according to the adjustment time sequence and polishing process stage, and form a dynamic optimization parameter sequence that includes parameter change trajectory, adjustment trigger conditions and corresponding equipment operation status.

6. The polishing process optimization method combining big data analysis according to claim 5, characterized in that, The calculation of the adjustment range for each process parameter to be adjusted is based on the degree of deviation, the gap between the predicted polishing effect and the expected effect, and the constraint requirements in the parameter coupling rules, including: An adjustment range calculation process is constructed, which includes steps for quantifying the impact of deviation, quantifying the effect gap, quantifying rule constraints, and comprehensive calculation. Each step works together to complete the adjustment range calculation. In the deviation impact quantification step, the degree of deviation between the actual executed value of the parameter in the real-time running data and the initial process parameter is converted into a quantitative index, which is used to reflect the magnitude of the impact of the deviation on the process execution. In the step of quantifying the effect gap, the difference between the predicted value of each indicator in the predicted polishing effect data and the preset expected effect indicator value is calculated, and the difference is converted into a standardized effect gap quantification value. The magnitude of the quantification value corresponds to the degree of significance of the effect gap. In the rule constraint quantification step, the adaptive constraint conditions related to the process parameter item to be adjusted are extracted from the process parameter coupling rule set, and the constraint conditions are converted into quantified constraint values. The quantified constraint values ​​are used to reflect the allowable range of parameter adjustment. The deviation impact quantification index, effect gap quantification value, and rule constraint quantification value are standardized based on a unified evaluation scale to obtain standardized deviation impact value, effect gap value, and rule constraint value. Set weighting coefficients for deviation impact value, effect gap value, and rule constraint value. The weighting coefficients are set based on the importance of each factor to parameter adjustment and historical adjustment experience. The standardized deviation impact value, effect gap value, and rule constraint value are multiplied by their corresponding weight coefficients to obtain the weighted evaluation results of each factor. In the comprehensive calculation step, the weighted evaluation results of each factor are summed to obtain the comprehensive evaluation value of parameter adjustment. The comprehensive evaluation value is used to reflect the necessity and adjustment space of parameter adjustment. An adjustment range mapping function is constructed, which takes the comprehensive evaluation value as input and the parameter adjustment range as output. The functional relationship is set based on the correspondence between the comprehensive evaluation value and the actual adjustment range in a large number of historical parameter adjustment process cases. The comprehensive evaluation value of parameter adjustment is input into the adjustment range mapping function to obtain the preliminary adjustment range value, which is the reference range of parameter adjustment; Based on the current value range of the process parameter to be adjusted, the initial adjustment range value is checked for boundary conditions to ensure that the adjusted parameter value is between the maximum and minimum allowable range of the process. Based on the boundary verification results, the initial adjustment range value is corrected. If the initial adjustment range value causes the parameter to exceed the boundary, the adjustment range is corrected to the difference or sum of the boundary value and the current parameter value. The corrected adjustment range value is then fine-tuned by referring to the historical parameter adjustment range data of similar workpieces to make the adjustment range more in line with the actual process execution requirements, and the final adjustment range of each process parameter item to be adjusted is determined.

7. The polishing process optimization method combining big data analysis according to claim 1, characterized in that, Based on the dynamically optimized parameter sequence, the system integrates equipment operation adaptation requirements, process execution timing, and quality control nodes to output a complete optimized polishing process solution, including: Extract parameter configuration information for each process stage from the dynamic optimization parameter sequence, divide the process execution units according to the order of polishing processes, and each process execution unit corresponds to a set of parameter configurations and a process execution stage; Analyze the parameter configuration information of each process execution unit, combine it with the equipment operation data in the multi-source data linkage set of the polishing process, retrieve the equipment operation records that match each parameter configuration from the historical equipment operation data, and extract the standard operation action sequence and parameter setting value from them; Based on the standard operating sequence and parameter settings, a set of equipment control instructions is generated for each process execution unit. The set of equipment control instructions includes the specified equipment operating mode, the standard operating sequence, and the parameter calibration target value. Based on the time nodes for parameter adjustment in the dynamic optimization parameter sequence and the order of process execution units, plan the process execution sequence and mark the start time, duration, and connection time of each process execution unit; Historical polishing quality control data is extracted from the multi-source data linkage set of the polishing process. Combined with the parameter change nodes in the dynamic optimization parameter sequence, the quality control nodes of each process execution unit are determined, and the control objects and control methods are marked. Configure the corresponding quality inspection program and inspection trigger conditions for each quality control node; Integrate the parameter configuration information, equipment operation adaptation requirements, process execution sequence and quality control nodes of each process execution unit to form a process execution unit plan. Each process execution unit plan corresponds to a complete process execution link. All process execution unit schemes are connected in sequence according to the process execution time to construct the main process execution flow, and the connection logic and data transmission requirements between process execution units are marked; equipment maintenance data corresponding to the dynamic optimization parameter sequence are extracted from the multi-source data linkage set of the polishing process, and combined with the main process execution flow, equipment pre-maintenance suggestions and process maintenance plans are formulated. Based on the main process flow, equipment pre-maintenance recommendations, and process maintenance plan, estimate the human, time, and material resources required for the polishing process and form a resource requirement list. Construct emergency response procedures during process execution, and define the triggering conditions, execution steps, and execution entities for possible equipment failures, parameter deviations, and quality anomalies. Integrate the main process execution flow, equipment operation adaptation requirements, quality control nodes, equipment maintenance suggestions, resource requirements list and emergency handling procedures to form a complete optimization plan for the polishing process, and output it in a standardized format.

8. The polishing process optimization method combining big data analysis according to claim 7, characterized in that, The emergency response procedures during the construction process clearly define the response measures and division of responsibilities for potential equipment failures, parameter deviations, and quality anomalies, including: Extract historical polishing process failure case data from the multi-source data linkage set of the polishing process, and statistically analyze the possible equipment failure types, parameter deviation types and quality anomaly types to form a failure anomaly type list; For each type of equipment failure, analyze the common causes, characteristics and impact on process execution, and extract effective failure handling steps based on historical experience. For each type of parameter deviation, the triggering conditions, changing trends, and impact on polishing effect are analyzed, and deviation correction measures are formulated in conjunction with the process parameter coupling rule set. For each type of quality anomaly, analyze the key nodes where the anomaly occurs, the propagation path, and the impact on the final polishing quality, and formulate anomaly containment measures and quality remediation plans. Set the response priority for each type of fault / abnormality. The response priority is set based on the degree of impact of the fault / abnormality on the process schedule, quality, and cost. Delineate the responsible departments and personnel for emergency response, and mark the scope of fault and anomaly handling and the responsibilities and authority of each responsible entity; Define an emergency response process, which includes an automated logical chain of fault / abnormal triggering, measure matching and execution, and feedback of processing results; Set emergency response time thresholds and mark the maximum allowable time for handling each type of fault or anomaly. The initial emergency handling rule base is generated by combining fault handling steps, deviation correction measures, anomaly containment measures, response priorities, execution entity definitions, emergency handling procedures, and time thresholds. The initial emergency handling rule base is then updated based on historical processing data, adding rules for special scenarios and correcting execution logic to generate an emergency handling rule base that includes handling solutions for all types of faults and anomalies.

9. The polishing process optimization method combining big data analysis according to claim 2, characterized in that, The analysis examines the cross-influence of process parameters across different parameter categories, extracting cross-category parameter adjustment correlation cases where changes in a process parameter in one category trigger adjustments in process parameters in other categories. Based on these cross-category parameter adjustment correlation cases, association rules describing the influence relationships between parameter categories are generated, including: From the multi-source data linkage set of the polishing process, select complete polishing process cases that contain records of changes in process parameter items of multiple parameter categories. The complete polishing process cases contain clear time nodes of parameter changes and corresponding adjustment records. The process parameters in each complete polishing process case are classified and labeled according to parameter category, and the parameter category to which each process parameter belongs and its functional position in process execution are marked. Construct a parameter change tracking map, with time as the horizontal axis and parameter category as the vertical axis, to mark the numerical changes of each process parameter at different time points, and intuitively present the parameter change trajectory; Identify the first change node of a process parameter item in a parameter category from the parameter change tracking map, and use the first change node as the starting point for cross-influence analysis; Track the changes in the values ​​of process parameters in other parameter categories after the starting point, and record the category, adjustment time, and adjustment range of the first process parameter item whose value is adjusted. Analyze the correlation between changes in initial parameters and subsequent parameter adjustments, extract cross-category parameter adjustment correlation cases directly caused by changes in initial parameters, and record the parameter category combinations, initial parameter change characteristics, subsequent parameter adjustment characteristics, and corresponding polishing scene information in the cross-category parameter adjustment correlation cases; Cross-category parameter adjustment association cases are classified, and case types are divided according to the combination of the starting parameter category and the affected parameter category. Each case type contains multiple similar cross-category parameter adjustment association cases. In each case type, the correspondence between the magnitude of change of the initial parameter and the magnitude of adjustment of the subsequent parameter is statistically analyzed, and the numerical ratio and change pattern between the magnitude of change of the initial parameter and the magnitude of adjustment of the subsequent parameter are calculated and recorded. Analyze the differences in the influence of cross-category parameters under different polishing scenarios within the same case type, extract the moderating effect of scenario features on the degree of influence, and label the influence weight of scenario factors; The common data patterns of cross-category parameter influence in each case type are statistically analyzed, including the influence delay time, the proportional relationship of adjustment magnitude, and the relationship of parameter change direction, and preliminary influence association rules are generated based on this. By integrating the preliminary influence association rules of all case types and supplementing the special adaptation items of influence in special scenarios, a cross-category parameter influence association rule set is generated. The cross-category parameter influence association rule set is used to describe the cross-influence relationship of process parameter items between different parameter categories.

10. A polishing process optimization system combining big data analysis, characterized in that, It includes a processor and a readable storage medium storing a program that, when executed by the processor, implements the polishing process optimization method combining big data analysis as described in any one of claims 1-9.

Images