A method and system for optimizing nanofilm preparation process
By establishing production data sets and optimizing the optimization space, the problem of difficult to accurately regulate nano-film preparation process parameters is solved, and the uniformity and performance of the film are improved.
Patent Information
- Application Number
- CN202411218901.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-02
- Publication Date
- 2024-12-13
- Estimated Expiration
- 2044-09-02
AI Technical Summary
The nano film preparation process parameters are difficult to accurately regulate, resulting in poor uniformity of the films obtained from the preparation.
By establishing production data sets, including public and encrypting data, data clustering and division, establishing basic and variation optimization space, performing key evaluation of process parameters and feature dimensionality reduction, combining fitness function for iterative optimization search, and optimizing process parameters.
It realizes precise control of process parameters, improves the uniformity and performance stability of the film, and breaks through the limitations of the traditional empirical function model.
Smart Images
Figure CN118732630B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field related to process data processing, and in particular to a method and system for optimizing a nano film preparation process. Background Art
[0002] Nanofilm refers to a thin film material formed by embedding components with a size of nanometers (1~100nm) in a matrix. It has the advantages of both traditional composite materials and modern nanomaterials, and has broad application prospects in many fields. For example, nano-silicon crystal film has the advantages of good thermal stability, strong light absorption ability, high doping effect, and high-temperature conductivity that can vary over a wide range. Therefore, it has broad application prospects in sensors, optoelectronic and electromagnetic devices and other thin-film microelectronic devices.
[0003] Nanofilms can also be used as optical device materials, microchip thermal conductive substrates and wiring materials, microelectronic packaging materials and solar cell materials. Process parameter optimization is a key link in improving the quality and production efficiency of nanofilm preparation. However, in the existing technology, the optimization of nanofilm preparation process parameters mostly relies on empirical function models, which leads to technical problems such as low production efficiency and poor quality of film preparation products.
[0004] In summary, in the prior art, there is a technical problem that the process parameters for preparing nanofilms are difficult to accurately control, resulting in poor uniformity of the prepared films. Summary of the invention
[0005] The present application provides a method and system for optimizing the nanofilm preparation process, aiming to solve the technical problem in the prior art that the nanofilm preparation process parameters are difficult to accurately control, resulting in poor uniformity of the prepared film.
[0006] In view of the above problems, the technical solution to implement this application is:
[0007] In a first aspect of the present application, a method for optimizing a nanofilm preparation process is provided, wherein the method comprises: establishing a production data set, wherein the production data set comprises a public production data set and an encrypted production data set, and each data in the production data set has a data correlation with a target prepared film;
[0008] The production data set is synchronized to the trusted processing terminal, and data clustering is performed according to data relevance and data credibility to establish a basic optimization space and a variation optimization space;
[0009] Sending an evaluation instruction of process parameters to the trusted processing end, performing a key evaluation of process parameters in the basic optimization space, and receiving the evaluation results fed back by the trusted processing end, performing feature dimensionality reduction, and establishing key process parameters;
[0010] Send a synchronization request to the trusted processing end, and obtain a marked basic optimization space according to the request response result, wherein the marked basic optimization space corresponds to the basic optimization space, and the marked basic optimization space stores public data and undisclosed numbers, and the undisclosed number is a unique mapping number of the encrypted data;
[0011] Establish a fitness function based on the required film characteristics and key process parameters, use the marked basic optimization space as the basic search space, and use the mutation optimization space as the additional search space. After setting the search probability, perform iterative optimization search based on the fitness function and the undisclosed number.
[0012] A solution space is established according to the results of iterative optimization search, and production control is performed using the solution space. Control feedback is received, and the solution space is updated using the control feedback results. Process optimization is completed based on the updated results.
[0013] A second aspect of the present application provides a system for optimizing a nanofilm preparation process, wherein the system comprises: a data set establishment unit, for establishing a production data set, wherein the production data set comprises a public production data set and an encrypted production data set, and each data in the production data set has a data association with a target film to be prepared;
[0014] A clustering division unit is used to synchronize the production data set to the trusted processing end, and to perform data clustering division according to data association and data credibility, and to establish a basic optimization space and a variation optimization space;
[0015] A feature dimension reduction unit, used to send an evaluation instruction of process parameters to the trusted processing end, perform a key evaluation of process parameters in the basic optimization space, and receive the evaluation result fed back by the trusted processing end, perform feature dimension reduction, and establish key process parameters;
[0016] The optimization space acquisition unit is used to send a synchronization request to the trusted processing end, and acquire the marked basic optimization space according to the request response result, wherein the marked basic optimization space corresponds to the basic optimization space, and the marked basic optimization space stores public data and undisclosed numbers, and the undisclosed numbers are unique mapping numbers of the encrypted data;
[0017] An optimization search unit is used to establish a fitness function based on the required film characteristics and key process parameters, and use the marked basic optimization space as the basic search space and the variation optimization space as the additional search space. After setting the search probability, it performs iterative optimization search based on the fitness function and the undisclosed number;
[0018] The production control unit is used to establish a solution space based on the results of iterative optimization search, and to perform production control with the solution space, and to receive control feedback, to update the solution space with the control feedback results, and to complete process optimization according to the updated results.
[0019] In summary, one or more technical solutions provided in the present application solve the technical problem that it is difficult to accurately control the process parameters of nano-film preparation, resulting in poor uniformity of the prepared film. By collecting parameters and performing parameter analysis with the assistance of artificial intelligence, combined with the determination of key process parameters, the optimization effect is guaranteed on the basis of reducing the amount of calculation, thereby making the constructed optimization effect more accurate, and through real-time monitoring of the linkage sensor, the precise execution of the process parameters is guaranteed, the preparation process is optimized, and the technical effect of improving the uniformity of the prepared film is improved. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] Figure 1 A possible schematic flow diagram of a method for optimizing a nanofilm preparation process is provided for this application;
[0021] Figure 2 A schematic flow diagram of possible feature dimensionality reduction in a method for optimizing a nanofilm preparation process is provided for this application;
[0022] Figure 3 A possible structural schematic diagram of a system for optimizing the nanofilm preparation process is provided for this application.
[0023] Explanation of the reference numerals: data set establishment unit M100, cluster division unit M200, feature dimension reduction unit M300, optimization space acquisition unit M400, optimization search unit M500, production control unit M600. DETAILED DESCRIPTION
[0024] Embodiment 1
[0025] The present application is described in detail below with reference to the accompanying drawings. Figure 1 As shown, the present application provides a method for optimizing a nanofilm preparation process, wherein the method comprises:
[0026] S1: establishing a production data set, wherein the production data set includes a public production data set and an encrypted production data set, and each data in the production data set has a data correlation with a target film to be prepared;
[0027] S2: Synchronize the production data set to the trusted processing end, perform data clustering according to data relevance and data credibility, and establish a basic optimization space and a variation optimization space;
[0028] As is known, the preparation process of nanofilms is often affected by many factors, including the selection of raw materials, film forming conditions (such as temperature, pressure, gas environment, etc.), deposition rate, precise control of film composition and structure, etc. Since the optimization of nanofilm preparation process parameters mostly relies on empirical function models, which mostly use single index mapping or multi-index mapping, and have simple analysis and processing logic, it is found in the long-term application process that it is difficult for empirical function models to accurately quantify the complex interactions between parameters, and it is difficult to effectively capture and deeply analyze the nonlinear effects in the nanofilm preparation process;
[0029] From the above analysis, we can see that there are many problems in the general nanofilm preparation process that need to be addressed, such as the difficulty in accurately controlling the process parameters, poor film uniformity, and insufficient repeatability. These problems restrict the practical application and large-scale production of high-performance nanofilms.
[0030] Based on this, while protecting the privacy of sensitive data, this application makes full use of multi-party production data for process optimization. On the one hand, a public production data set is established in the production data set, and on the other hand, an encrypted production data set is established in the production data set, and the optimization exploration of encrypted data is performed in the variation optimization space against the undisclosed number, so as to more comprehensively analyze the mutual influence relationship between parameters, improve the accuracy and scope of application of process parameter optimization, and thus ensure the security of data;
[0031] On this basis, the optimization space is reasonably divided, the key process parameters that play a decisive role in the performance of the film are effectively identified, and the fitness function is designed to quantify the matching degree between the film characteristics and the process parameters. Under the search strategy that takes into account the basic and variation spaces, the optimal solution space is explored in an iterative optimization manner, and finally the precise control and continuous optimization of the nano film preparation process are achieved. The changes in the production environment are flexibly responded to and the nano film preparation process is continuously optimized, breaking through the limitations of the traditional empirical function model in dealing with the interaction of complex process parameters, and effectively improving the performance stability and product quality of the nano film.
[0032] The method of the present application comprises as a whole: collecting all production data related to the preparation of nanofilms, wherein the production data set can be divided into two parts: public production and encrypted production, wherein the public production data set is raw data that can be used directly, and the encrypted production data set contains important data that has been encrypted to protect trade secrets or sensitive information; each data in the production data set is associated with a performance index of a specific target prepared film, that is, each data point has its corresponding data correlation, so that the influence of each data in the production data set on the film performance can be clarified in subsequent analysis;
[0033] The constructed production data set is securely synchronized to a pre-set trusted processing end. The trusted processing end, i.e., a trusted third party, has the right to read all data and has sufficient security measures to ensure that data privacy is not leaked and that effective data processing and analysis can be performed. The multiple terminals connected to the trusted processing end only have the right to read the public production data set and the encrypted production data set corresponding to their own part. On the trusted processing end, the entire data set is clustered and divided based on data relevance and data credibility to form a basic optimization space and a variation optimization space. The basic optimization space mainly covers the part that has a significant impact on film performance and has relatively reliable data, and the variation optimization space is a space that contains potential influencing factors or newly emerging process parameters.
[0034] The above steps build and analyze production data sets, use trusted processing terminals to perform safe and efficient parameter evaluation and optimization space division, and provide a basis for subsequent processing.
[0035] S3: Sending an evaluation instruction of process parameters to the trusted processing end, performing a key evaluation of process parameters in the basic optimization space, and receiving the evaluation results fed back by the trusted processing end, performing feature dimension reduction, and establishing key process parameters;
[0036] Furthermore, the evaluation results fed back by the trusted processing end are received and feature dimension reduction is performed, such as Figure 2 As shown, the application method includes:
[0037] S31: converting the evaluation results into parameter proportions to establish a parameter proportion data set;
[0038] S32: configuring an adaptive screening interval of a proportion, sorting the proportion data set in sequence, selecting an interval of the sorting result by using the adaptive screening interval, and determining a parameter set within the selection range;
[0039] S33: performing a mutation evaluation of a reduction in the proportion of the parameter set, and determining a critical parameter according to the evaluation result;
[0040] S34: Perform feature dimensionality reduction according to the critical parameters and the order sorting results.
[0041] The method of the present application comprises as a whole: sending an evaluation instruction on process parameters to a trusted processing end, allowing the trusted processing end to perform a key evaluation on the importance of each process parameter in a basic optimization space, and the trusted processing end will evaluate each process parameter in the production data set one by one according to a preset evaluation model and algorithm, and analyze the influence weight of each process parameter on film performance (such as principal component analysis PCA, etc.), including quantitative evaluation of the degree of influence of the parameter on film performance, the interaction relationship between parameters, and the sensitivity of parameter changes to film quality; receiving the evaluation results of the process parameters from the trusted processing end, performing feature dimensionality reduction processing according to the evaluation results, thereby eliminating non-critical dimensions and redundant information, and refining key process parameters for nanofilm preparation;
[0042] Receive the evaluation results fed back by the trusted processing end and perform feature dimension reduction. Specifically, receive the feedback results of the key evaluation of the process parameters from the trusted processing end, and the feedback results include each process parameter and the degree of influence of each process parameter on the film performance; convert the evaluation results of each process parameter in the production data set into a proportion of the overall influence proportion, construct a parameter proportion data set, and intuitively display the importance ranking of each parameter on the film performance;
[0043] According to the actual situation and optimization requirements, set an adaptive filtering interval, for example, only focus on the parameters whose proportions fall within the filtering interval; sort the data set according to the parameter proportions, so that the parameters with higher proportions are ranked first; in the sorted parameter proportion data set, according to the adaptive filtering interval, select the parameters whose proportions are within the interval, and confirm the parameter set within the selected range;
[0044] For example: A: 30%, B: 25%, C: 20%, D: 10%, E: 5%, F: 3%, G: 2%, ...; the adaptive screening interval can be set to 70%-85%. At this time, C (30%+25%+20%=75%, 75% belongs to 70%-85%) and D (30%+25%+20%+10%=85%, 85% belongs to 70%-85%) fall into the screening interval; the examples given in the examples are few and are only used as examples. In actual application, multiple parameters are included for independent selection;
[0045] Perform mutation analysis on the determined parameter set. Mutation reduction refers to artificially reducing the influence of one or several parameters, observing the changes in film performance, and evaluating the absolute influence and relative stability of the corresponding adjusted parameters; based on the results of mutation reduction evaluation, identify the critical parameters that have a key impact on film performance and may cause a sharp drop in performance once changed
[0046] According to the results of critical parameters and order sorting, the critical parameters and non-critical parameters with a high proportion are retained first, and the parameters with a low proportion and small non-critical influence are removed or weakened; the dimension reduction model is constructed using the screened core parameters to simplify the original high-dimensional process parameter space into a low-dimensional feature space, thereby achieving feature dimension reduction and focusing on optimizing the most critical parameters of the nanofilm preparation process;
[0047] The above steps receive the evaluation results from the trusted processing end, and perform a series of parameter ratio conversion, sorting, screening and dimensionality reduction operations to finally extract the most influential process parameters for subsequent process optimization iterations.
[0048] S4: Send a synchronization request to the trusted processing end, and obtain a marked basic optimization space according to the request response result, wherein the marked basic optimization space corresponds to the basic optimization space, and the marked basic optimization space stores public data and undisclosed numbers, and the undisclosed numbers are unique mapping numbers of encrypted data;
[0049] The overall method of the present application includes: sending a synchronization request to a trusted processing end, requesting to obtain a labeled basic optimization space after data association and credibility analysis; receiving a response result from the trusted processing end, obtaining a labeled basic optimization space containing a unique mapping number of public data and encrypted data (i.e., an undisclosed number), wherein the undisclosed number refers to a mapping number of undisclosed data (the number is public, and the encrypted production data corresponding to the undisclosed number is not public), and correspondingly, the public data can be directly used for further analysis, while the undisclosed number is used for subsequent processing and optimization work without exposing the original encrypted data, providing a basis for subsequent processing;
[0050] S5: Establish a fitness function based on the required film characteristics and key process parameters, and use the marked basic optimization space as the basic search space and the mutation optimization space as the additional search space. After setting the search probability, perform iterative optimization search based on the fitness function and the undisclosed number;
[0051] Furthermore, the fitness function is established based on the required film characteristics and key process parameters, and the marked basic optimization space is used as the basic search space, and the variation optimization space is used as the additional search space. After setting the search probability, an iterative optimization search is performed based on the fitness function and the undisclosed number. The method of the present application includes:
[0052] S51: In the marked basic optimization space, a basic solution set space is established using public data;
[0053] S52: Perform search determination of the basic search space and the additional search space by using the search probability;
[0054] S53: If the search result is a basic search space, then randomly select parameters in the basic search space, perform fine-tuning and expansion, and perform fitness analysis of the fine-tuning and expansion result and the basic solution set space through a fitness function, and update the basic solution set space according to the analysis result to complete an optimization search;
[0055] S54: if the search determination result is an additional search space, an additional selected command is generated and sent to the trusted processing end, and processing feedback from the trusted processing end is received, wherein the processing feedback is the additional expanded feedback result, and fitness analysis of the processing feedback and the basic solution set space is performed according to the fitness function, and the basic solution set space is updated according to the analysis result, and an optimization search is completed;
[0056] S55: Complete iterative optimization search based on multiple rounds of search.
[0057] The overall method of the present application includes: combining the required characteristics of the target film and the key process parameters extracted in the previous step, formulating a fitness function, wherein the fitness function can measure the degree of fit between the process parameter combination and the desired film characteristics, and provide an evaluation standard for subsequent optimization; using the marked basic optimization space as the basic search area, and the variation optimization space as the additional search range, and starting the iterative optimization search algorithm based on the fitness function and the undisclosed number under the premise of setting the search probability; in each iteration process, selecting the process parameter combination to be optimized in the optimization space according to the search probability, and indirectly operating the encrypted data through the undisclosed number, and calculating the corresponding fitness score, so as to find the optimal or suboptimal process parameter configuration;
[0058] Establish a fitness function based on the required film characteristics and key process parameters, and use the marked basic optimization space as the basic search space and the variant optimization space as the additional search space. After setting the search probability, perform iterative optimization search based on the fitness function and the undisclosed number. Specifically, design a fitness function for quantitatively evaluating film performance based on the required film characteristic indicators and key process parameters, so that the closer the parameter combination is to the ideal value, the higher the fitness function value; use the public data in the marked basic optimization space to preliminarily construct the basic solution set space, which contains all feasible process parameter combinations that are currently known and meet certain conditions;
[0059] Set the search probability to decide whether to explore in the basic search space or the variant (additional) search space in each iteration; make a random judgment based on the preset probability model to determine whether the current search is conducted in the basic search space or the additional search space. Furthermore, in the first scenario, if the judgment result points to the basic search space, randomly select an existing parameter combination in the space, and make a small perturbation or adjustment (i.e., fine-tune and expand) to the selected existing parameter combination; apply the fitness function to evaluate the performance of the fine-tuned parameter combination and compare it with other combinations in the current basic solution set space; if the newly obtained parameter combination has a higher fitness, then incorporate the fine-tuned parameter combination into the basic solution set space, replace the original combination with a lower fitness, and complete an optimization iteration in the basic search space;
[0060] In the second scenario, if the determination result points to an additional search space, a corresponding additional selected command is generated, and the additional selected command is sent to the trusted processing end to request an additional expansion operation related to the encrypted data; the trusted processing end receives the feedback result after processing based on the undisclosed number, uses the fitness function to evaluate the received processing feedback result, and compares it with the existing combinations in the basic solution set space; if the newly obtained parameter combination has a higher fitness, it is also added to the basic solution set space to replace the original combination with a lower fitness, completing an optimization iteration of the additional search space;
[0061] Repeat the process of the first scenario and the second scenario, perform multiple rounds of search iterations, and continuously optimize the process parameter combinations in the basic solution set space until a preset stopping criterion is reached, such as reaching a certain fitness threshold, satisfying a certain number of iterations or other convergence conditions; finally, after multiple rounds of iterative optimization, a set of optimal process parameter combinations determined in the basic solution set space can be used to guide actual thin film production process control.
[0062] Furthermore, the search determination of the basic search space and the additional search space by the search probability, the method of the present application also includes:
[0063] S521: Record the fitness value of the basic search space, and establish a penalty factor according to the fitness value and the number of consecutive selections of the basic search space;
[0064] S522: adding the penalty factor to the search probability, and performing search decision constraints on the basic search space and the additional search space;
[0065] S523: When the search determination result is an additional search space, the penalty factor is reset, and the search determination is completed according to the penalty factor.
[0066] The search probability is used to make search decisions for the basic search space and the additional search space. Specifically, after each optimization of the basic search space is completed, the fitness value of the current optimal solution (or multiple solutions that are satisfied by parallel processing, which can be selected autonomously through the corresponding hardware configuration) is recorded; according to the number of consecutive selections in the basic search space, a dynamically changing penalty factor is designed. For example, if the basic search space is selected for multiple consecutive times but the fitness is not improved much, the penalty factor can be gradually increased to reduce the probability of subsequent selection of this space; before making a search decision, the current penalty factor is merged into the selection probability of the basic search space. Therefore, when the continuous exploration effect of the basic search space is not good, it is more inclined to explore the additional search space to find the potential global optimal solution;
[0067] Combined with the updated search probability (including the penalty factor), a random judgment is made again to decide whether to enter the additional search space; when the search judgment result is still the basic search space, the optimization continues according to the previous basic search space optimization process; when the search judgment result shows that it is an additional search space, not only the above additional search space optimization process is executed, but also the penalty factor is reset to the initial value, so that the next round of search judgment will not be affected by the previous number of consecutive selections;
[0068] After a round of optimization in the additional search space, no matter what the result is, return to step 6 to re-evaluate the fitness value of the basic search space, update the penalty factor, and proceed to the next round of search judgment and optimization process; the continuous cycle iteration will continue until the pre-set termination condition is met, ensuring that while considering the search space switching, the optimal solution in the basic search space and the additional search space is fully explored.
[0069] Furthermore, if the search determination result is a basic search space, parameters in the basic search space are randomly selected for fine-tuning and expansion. The method of the present application includes:
[0070] S531: If the randomly selected result is encrypted data, the selected undisclosed number is fed back to the trusted processing end, and the encrypted fine-tuning expansion result of the trusted processing end is received;
[0071] S532: Update the basic solution set space with the encrypted fine-tuning expansion result.
[0072] If the search result is a basic search space, then parameters in the basic search space are randomly selected for fine-tuning and expansion. Specifically, after determining that the search space is the basic search space, a set of parameters is randomly selected from the public data in the basic search space for fine-tuning and expansion, assuming that the basic search space contains both public data and encrypted data mapping numbers;
[0073] Check whether the selected parameters contain an undisclosed number (the mapping number corresponding to the encrypted production data set); if the randomly selected parameters contain an undisclosed number, send the number to the trusted processing end; wait for the trusted processing end to fine-tune and expand the encrypted data, and receive the returned encrypted fine-tune and expansion results. Limitedly, during this process, the trusted processing end will not leak any original encrypted data, but will calculate and return the adjustment results in encrypted form through a secure protocol;
[0074] Parse the received encrypted fine-tuning and expansion results and regard them as a correction or optimization of the original parameters; apply the encrypted fine-tuning and expansion results to the original parameter combination to replace or supplement the corresponding part of the original encrypted data; use the fitness function to evaluate the conformity of the updated parameter combination to the target film characteristics; if the encrypted fine-tuning and expansion parameter combination has a higher fitness, add it to the basic solution set space to replace the original combination with lower fitness, or retain it as a new candidate solution;
[0075] Even when facing a basic search space containing encrypted data, the above steps can safely perform parameter fine-tuning and optimization with the help of a trusted processing terminal, ultimately achieving effective updating and expansion of the basic solution space.
[0076] S6: Establish a solution space according to the results of the iterative optimization search, and use the solution space to perform production control, and receive control feedback, update the solution space with the control feedback result, and complete the process optimization according to the updated result.
[0077] Furthermore, the method of updating the solution space by controlling the feedback result includes:
[0078] S61: Determine whether the deviation value of the control feedback result meets a preset threshold;
[0079] S62: If the preset threshold is met, an early warning message is generated, and a maintenance early warning of the equipment is performed based on the early warning message. After the equipment maintenance is performed, the early warning message is eliminated.
[0080] The method of the present application comprises as a whole: creating a solution space according to the results of iterative optimization search, the solution space including the optimal or relatively optimal process parameter configuration set, and the process parameter configuration set corresponding to the solution space can be used to guide the actual nano film preparation process; applying the optimal parameter combination in the solution space to the production line, controlling the nano film preparation process, and adjusting the equipment parameters in real time to achieve process optimization;
[0081] After that, it also includes, after executing the new process parameter configuration, receiving the control feedback data on the production line in real time, the control feedback data reflects the gap between the actual film performance produced and the expected target; updating and adjusting the solution space according to the control feedback results, the optimized solution space will reflect the latest process parameter optimization status, and enter the iterative optimization search process again until the preset process optimization target or convergence condition is reached;
[0082] Update the solution space with the control feedback results. Specifically, continuously collect control feedback data during system operation, such as key information such as the working status and performance indicators of the equipment; compare the currently acquired control feedback results with the preset ideal values or standard values, calculate the actual deviation values of each parameter, and compare the calculated deviation values with the preset threshold values to check whether any parameter deviation value exceeds the allowable range;
[0083] If the deviation value of one or more parameters does exceed the preset threshold, it is determined that there is a potential problem or failure risk; based on the specific parameters that exceed the threshold and the degree of deviation, detailed warning information is automatically created, including but not limited to the warning level, the equipment components involved, possible cause analysis and recommended countermeasures; the generated warning information is sent to relevant personnel in a timely manner to start the equipment maintenance warning process; after receiving the warning information, the equipment is inspected and repaired as necessary according to the warning instructions;
[0084] After completing equipment maintenance, the control feedback results are collected again to check whether the relevant parameters have returned to the normal range; after confirming that the equipment working state has returned to normal and the deviation values of all parameters are lower than the preset thresholds, the warning information is eliminated to ensure that production activities can continue safely and efficiently. Real-time monitoring of control feedback results, deviation warnings and equipment maintenance management based on warning information are realized, which effectively ensures the overall stability of the nanofilm preparation process and further improves the uniformity of the prepared films.
[0085] Furthermore, the method of determining whether the deviation value of the control feedback result meets a preset threshold value includes:
[0086] S611: If the preset threshold is not met, a common offset interval is established based on the control feedback result;
[0087] S612: Perform compensation optimization of control parameters in the solution space through the common offset interval, and update the solution space with the compensation optimization result.
[0088] Determine whether the deviation value of the control feedback result meets the preset threshold value. Specifically, obtain the actual control feedback result and compare it with the preset target or ideal value to calculate the deviation value of each control parameter; when it is found that the deviation value of some control parameters does not meet the preset threshold requirement, perform statistical analysis on the control parameters that do not meet the threshold, find the common deviation rules and trends of multiple control parameters that do not meet the threshold, and thus define a common deviation interval covering these deviation characteristics;
[0089] Based on the characteristics of the common offset interval, design a suitable compensation model or algorithm for parameter compensation optimization in the solution space; use optimization algorithms (such as gradient descent, genetic algorithm, particle swarm optimization, etc.) to search for the optimal control parameter compensation value in the solution space, so that the compensated parameter value can offset or reduce the existing offset to the greatest extent; apply the best compensation value obtained by compensation optimization to the original solution space, and update the corresponding control parameter settings to optimize the overall control behavior and performance;
[0090] When the control feedback results deviate greatly in the above steps, a common deviation interval is constructed by analyzing the deviation characteristics, and on this basis, the optimal compensation parameters are found in the solution space, in order to improve the control system performance through compensation optimization, and finally achieve effective update of the solution space and improvement of the overall performance.
[0091] In summary, the beneficial effects of the embodiments of the present application are:
[0092] 1. By integrating production data sets, we can effectively utilize global data information while protecting data privacy, concentrate data resources from all aspects for deep mining and optimization, simplify the optimization process, focus on the key parameters that truly affect film performance, improve optimization efficiency, quickly lock in key process parameters and the best combination, and improve the accuracy of process optimization.
[0093] 2. Establish a trusted processing terminal so that participants can collaborate and optimize while maintaining data privacy, break through the limitations of data silos, and facilitate cross-institutional process improvements.
[0094] 3. Through iterative optimization search algorithms, combined with real-time control feedback during the production process, the process parameters can be dynamically adjusted and optimized, and online adjustment of the process parameters can be achieved, thereby improving process stability and consistency of film quality.
[0095] 4. The fitness value of the basic search space is recorded, and a penalty factor is established according to the fitness value and the number of consecutive selections of the basic search space; the penalty factor is added to the search probability, and the search decision constraints of the basic search space and the additional search space are performed; when the search decision result is the additional search space, the penalty factor is reset, and the search decision is completed according to the penalty factor, ensuring that the optimal solution in the basic search space and the additional search space is fully explored while considering the search space switching.
[0096] Embodiment 2
[0097] Based on the same inventive concept as the method for optimizing the nanofilm preparation process in the aforementioned embodiment, Figure 3 As shown, an embodiment of the present application provides a system for optimizing a nanofilm preparation process, wherein the system comprises:
[0098] A data set establishing unit M100 is used to establish a production data set, wherein the production data set includes a public production data set and an encrypted production data set, and each data in the production data set has a data correlation with a target film to be prepared;
[0099] The clustering division unit M200 is used to synchronize the production data set to the trusted processing end, and to perform data clustering division according to data relevance and data credibility, and to establish a basic optimization space and a variation optimization space;
[0100] The feature dimension reduction unit M300 is used to send an evaluation instruction of the process parameters to the trusted processing end, perform a key evaluation of the process parameters in the basic optimization space, and receive the evaluation results fed back by the trusted processing end, perform feature dimension reduction, and establish key process parameters;
[0101] The optimization space acquisition unit M400 is used to send a synchronization request to the trusted processing end, and obtain the marked basic optimization space according to the request response result, wherein the marked basic optimization space corresponds to the basic optimization space, and the marked basic optimization space stores public data and undisclosed numbers, and the undisclosed numbers are unique mapping numbers of encrypted data;
[0102] The optimization search unit M500 is used to establish a fitness function based on the required film characteristics and key process parameters, and use the marked basic optimization space as the basic search space and the variation optimization space as the additional search space. After setting the search probability, it performs iterative optimization search based on the fitness function and the undisclosed number;
[0103] The production control unit M600 is used to establish a solution space according to the iterative optimization search results, and to perform production control with the solution space, and to receive control feedback, to update the solution space with the control feedback result, and to complete process optimization according to the updated result.
[0104] Furthermore, the optimization search unit M500 is used to perform the following method:
[0105] In the marked basic optimization space, the basic solution set space is established with public data;
[0106] Perform search judgment of the basic search space and the additional search space through the search probability;
[0107] If the search result is the basic search space, the parameters in the basic search space are randomly selected for fine-tuning and expansion, and the fitness analysis of the fine-tuning and expansion results and the basic solution set space is performed through the fitness function. The basic solution set space is updated according to the analysis results to complete an optimization search;
[0108] If the search result is an additional search space, an additional selected command is generated and sent to the trusted processing end, and processing feedback from the trusted processing end is received, wherein the processing feedback is the additional expanded feedback result, and fitness analysis of the processing feedback and the basic solution set space is performed according to the fitness function, and the basic solution set space is updated according to the analysis result to complete an optimization search;
[0109] Iterative optimization search is completed based on multiple rounds of search.
[0110] Furthermore, the optimization search unit M500 is also used to perform the following method:
[0111] Recording the fitness value of the basic search space, and establishing a penalty factor according to the fitness value and the number of consecutive selections of the basic search space;
[0112] Adding the penalty factor to the search probability to perform search decision constraints on the basic search space and the additional search space;
[0113] When the search decision result is an additional search space, the penalty factor is reset, and the search decision is completed according to the penalty factor.
[0114] Furthermore, the optimization search unit M500 is also used to perform the following method:
[0115] If the randomly selected result is encrypted data, the selected undisclosed number is fed back to the trusted processing end, and the encrypted fine-tuning and expansion result of the trusted processing end is received;
[0116] The basic solution set space is updated with the encrypted fine-tuning expansion result.
[0117] Furthermore, the feature dimension reduction unit M300 is used to perform the following method:
[0118] Convert the evaluation results into parameter proportions and establish a parameter proportion data set;
[0119] Configure an adaptive screening interval of the proportion, sort the proportion data set in sequence, select the interval of the sequential sorting result by the adaptive screening interval, and determine the parameter set within the selection range;
[0120] Performing a mutation evaluation of the parameter set to determine a critical parameter based on the evaluation result;
[0121] Feature dimensionality reduction is performed based on critical parameters and order sorting results.
[0122] Furthermore, the production control unit M600 is used to perform the following method:
[0123] Determining whether the deviation value of the control feedback result meets a preset threshold;
[0124] If the preset threshold is met, an early warning message is generated, and a maintenance early warning of the equipment is performed based on the early warning message. After the equipment maintenance is performed, the early warning message is eliminated.
[0125] Furthermore, the production control unit M600 is also used to execute the following method:
[0126] If the preset threshold is not met, a common offset interval is established based on the control feedback result;
[0127] Compensation optimization of control parameters in the solution space is performed through the common offset interval, and the solution space is updated with the compensation optimization result.
[0128] In summary, any step can be stored as a computer instruction or program in an unlimited computer memory, and can be called and recognized by an unlimited computer processor, and no unnecessary restrictions are made here.
[0129] Furthermore, the above-mentioned technical scheme only reflects the preferred technical scheme of the technical scheme of the embodiment of the present application. Some changes that may be made to some parts thereof by technicians in this technical field all reflect the novel principles of the embodiment of the present application. Obviously, technicians in this field can make various changes and modifications to the present application without departing from the scope of the present application.
Claims
1. A method for optimizing the preparation process of nanofilms, characterized in that: The method comprises: Establishing a production data set, wherein the production data set includes a public production data set and an encrypted production data set, and each data in the production data set has a data correlation with a target film to be prepared; The production data set is synchronized to the trusted processing terminal, and data clustering is performed according to data relevance and data credibility to establish a basic optimization space and a variation optimization space; Sending an evaluation instruction of process parameters to the trusted processing end, performing a key evaluation of process parameters in the basic optimization space, and receiving the evaluation results fed back by the trusted processing end, performing feature dimensionality reduction, and establishing key process parameters; Send a synchronization request to the trusted processing end, and obtain a marked basic optimization space according to the request response result, wherein the marked basic optimization space corresponds to the basic optimization space, and the marked basic optimization space stores public data and undisclosed numbers, and the undisclosed number is a unique mapping number of the encrypted data; Establish a fitness function based on the required film characteristics and key process parameters, use the marked basic optimization space as the basic search space, and use the mutation optimization space as the additional search space. After setting the search probability, perform iterative optimization search based on the fitness function and the undisclosed number. Establishing a solution space according to the results of iterative optimization search, and using the solution space to perform production control, and receiving control feedback, updating the solution space with the control feedback results, and completing process optimization according to the updated results; The method of establishing a fitness function based on the required film characteristics and key process parameters, taking the marked basic optimization space as the basic search space, taking the variation optimization space as the additional search space, and performing iterative optimization search based on the fitness function and the undisclosed number after setting the search probability includes: In the marked basic optimization space, the basic solution set space is established with public data; Perform search judgment of the basic search space and the additional search space through the search probability; If the search result is the basic search space, the parameters in the basic search space are randomly selected for fine-tuning and expansion, and the fitness analysis of the fine-tuning and expansion results and the basic solution set space is performed through the fitness function. The basic solution set space is updated according to the analysis results to complete an optimization search; If the search result is an additional search space, an additional selected command is generated and sent to the trusted processing end, and processing feedback from the trusted processing end is received, wherein the processing feedback is the additional expanded feedback result, and fitness analysis of the processing feedback and the basic solution set space is performed according to the fitness function, and the basic solution set space is updated according to the analysis result to complete an optimization search; Complete iterative optimization search based on multiple rounds of search; If the search result is a basic search space, then parameters in the basic search space are randomly selected for fine-tuning and expansion, including: If the randomly selected result is encrypted data, the selected undisclosed number is fed back to the trusted processing end, and the encrypted fine-tuning and expansion result of the trusted processing end is received; The basic solution set space is updated with the encrypted fine-tuning expansion result.
2. The method according to claim 1, characterized in that The step of performing search determination of the basic search space and the additional search space by using the search probability further includes: Recording the fitness value of the basic search space, and establishing a penalty factor according to the fitness value and the number of consecutive selections of the basic search space; Adding the penalty factor to the search probability to perform search decision constraints on the basic search space and the additional search space; When the search decision result is an additional search space, the penalty factor is reset, and the search decision is completed according to the penalty factor.
3. The method according to claim 1, characterized in that Receive the evaluation results fed back by the trusted processing end and perform feature dimension reduction, including: Convert the evaluation results into parameter proportions and establish a parameter proportion data set; Configure an adaptive screening interval of the proportion, sort the proportion data set in sequence, select the interval of the sequential sorting result by the adaptive screening interval, and determine the parameter set within the selection range; Performing a mutation evaluation of the parameter set to determine a critical parameter based on the evaluation result; Feature dimensionality reduction is performed based on critical parameters and order sorting results.
4. The method according to claim 1, characterized in that The updating of the solution space by using the control feedback result includes: Determining whether the deviation value of the control feedback result meets a preset threshold; If the preset threshold is met, an early warning message is generated, and a maintenance early warning of the equipment is performed based on the early warning message. After the equipment maintenance is performed, the early warning message is eliminated.
5. The method according to claim 4, characterized in that The determining whether the deviation value of the control feedback result meets a preset threshold value includes: If the preset threshold is not met, a common offset interval is established based on the control feedback result; Compensation optimization of control parameters in the solution space is performed through the common offset interval, and the solution space is updated with the compensation optimization result.
6. A system for optimizing nanofilm preparation process, characterized in that: For executing the method according to any one of claims 1 to 5, the system comprises: A data set establishing unit, used for establishing a production data set, wherein the production data set includes a public production data set and an encrypted production data set, and each data in the production data set has a data correlation with a target film to be prepared; A clustering division unit is used to synchronize the production data set to the trusted processing end, and to perform data clustering division according to data association and data credibility, and to establish a basic optimization space and a variation optimization space; A feature dimension reduction unit, used to send an evaluation instruction of process parameters to the trusted processing end, perform a key evaluation of process parameters in the basic optimization space, and receive the evaluation result fed back by the trusted processing end, perform feature dimension reduction, and establish key process parameters; The optimization space acquisition unit is used to send a synchronization request to the trusted processing end, and acquire the marked basic optimization space according to the request response result, wherein the marked basic optimization space corresponds to the basic optimization space, and the marked basic optimization space stores public data and undisclosed numbers, and the undisclosed numbers are unique mapping numbers of the encrypted data; An optimization search unit is used to establish a fitness function based on the required film characteristics and key process parameters, and use the marked basic optimization space as the basic search space and the variation optimization space as the additional search space. After setting the search probability, it performs iterative optimization search based on the fitness function and the undisclosed number; The production control unit is used to establish a solution space based on the results of iterative optimization search, and to perform production control with the solution space, and to receive control feedback, to update the solution space with the control feedback result, and to complete process optimization according to the updated result.