Dimethyl sulfoxide production optimization method and system based on deep neural network
Patent Information
- Application Number
- CN202510829083.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-20
- Publication Date
- 2025-10-17
Smart Images
Figure CN120806596A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of intelligent production of dimethyl sulfoxide, in particular to a dimethyl sulfoxide production optimization method and system based on a deep neural network. BACKGROUND
[0002] In the existing dimethyl sulfoxide production process, different batches of raw materials are usually controlled by a unified reaction based on fixed process parameters, lacking dynamic perception and personalized analysis of batch process characteristics, making it difficult to effectively match the actual reaction characteristics of each batch of raw materials with the optimal process requirements.
[0003] The conventional process control method mainly relies on experience parameter setting and manual intervention, lacking dynamic feedback and self-adaptive adjustment mechanism based on real-time monitoring data, resulting in large product yield fluctuation, high energy consumption level and insufficient process stability, which is difficult to support high-precision and high-efficiency dimethyl sulfoxide production requirements.
[0004] At the data processing level, the existing technology lacks the ability of standardization and normalization processing and deep fusion modeling for multi-source process data such as reaction temperature, reaction pressure, catalyst concentration, raw material flow and energy consumption indicators, making it difficult to achieve accurate evaluation of process state, deep analysis of reaction mechanism and dynamic prediction of process trend.
[0005] At the same time, process parameter optimization mainly relies on static rule setting or offline summary based on historical experience, failing to combine real-time monitoring data trends, process deviation degrees and abnormal fluctuation characteristics for comprehensive analysis and dynamic adjustment, limiting the precision of process optimization, the stability of production process and the consistency of product quality, and making it difficult to meet the intelligent, precise and efficient dimethyl sulfoxide production process control requirements.
[0006] Therefore, in view of the deficiencies of the prior art, a dimethyl sulfoxide production optimization method and system based on a deep neural network are needed. SUMMARY
[0007] Technical problems solved
[0008] In view of the deficiencies of the prior art, the present application provides a dimethyl sulfoxide production optimization method and system based on a deep neural network, which solves the problem that the traditional dimethyl sulfoxide production process relies on experience rules and simple linear control methods, which cannot fully capture the complex process characteristics of nonlinearity and strong coupling, resulting in large fluctuations in process parameters and making it difficult to achieve real-time and accurate regulation of the production process.
[0009] Technical scheme
[0010] To achieve the above objectives, the present invention is implemented through the following technical solutions: a dimethyl sulfoxide production optimization method and system based on deep neural networks, including S1, deploying multiple types of detection terminals to collect process characteristic data, production environment data and processing process data of the current batch, and standardizing and normalizing the collected process characteristic data, production environment data and processing process data to construct a standardized data set; S2, based on the standardized data set and combined with the characteristics of each process parameter of historical batches, comprehensively evaluate the nonlinear coupling relationship between different process parameters, generate process parameter interaction values for the current batch, and optimize the subsequent data collection frequency and dynamic adjustment strategy priority according to the process parameter interaction values; S3, based on the standardized data set, perform process sensitivity analysis on each process parameter in the dimethyl sulfoxide production process under the characteristics of the current batch, and then update the adjustment amplitude range and dynamic correction strategy of each key process parameter according to the process sensitivity analysis results; S4, using the current batch process parameter interaction value and process sensitivity analysis results as input, comprehensively analyze the comprehensive dynamic matching degree of the current batch under nonlinear process characteristics and strong coupling parameter characteristics, and generate optimal process adjustment optimization suggestions based on the comprehensive dynamic matching degree analysis results.
[0011] Furthermore, the specific steps of deploying multiple types of detection terminals to collect process characteristic data, production environment data and processing process data of the current batch are as follows: collecting process characteristic data including raw material ratio, reaction temperature curve, reaction pressure curve, catalyst concentration, stirring rate change, raw material flow fluctuation, reaction time, product yield, product purity and energy consumption index through production process detection equipment, and recording the measurement unit, measurement method, detection accuracy, unit energy consumption change, detection equipment model, ratio of historical measurement mean to standard deviation, degree of deviation of current process characteristics from historical mean, instantaneous fluctuation amplitude and ambient temperature, humidity and Air pressure conditions; collect production environment data including the temperature, pressure, humidity, gas flow rate, heat exchange efficiency, equipment operation time, raw material consumption rate and condensation efficiency around the reactor through process environment monitoring equipment, and record the collection frequency, process operation frequency, change rate, real-time collection value, resource consumption number, monitoring cycle, historical monthly average and standard deviation, current process section number and production batch number of each process parameter; record the processing data including the total number of collected process parameters, timestamp of each collected data, sampling interval, measurement time, collection data completeness rate, historical collection average, unit energy consumption change average, batch data coverage rate and environmental monitoring synchronization rate.
[0012] Further, the collected process characteristic data, production environment data and processing process data are standardized and normalized for preprocessing, and the specific steps for constructing the standardized data set are as follows: through a unified time synchronization mechanism, the time stamp of all process characteristic data and production environment data is standardized, and the time alignment success rate and time drift average of each production batch data are recorded; the null value, invalid extreme value, format abnormal field and measurement error out-of-limit value in the process characteristic data, production environment data and processing process data are cleaned, and the missing rate, abnormal proportion and rejection ratio of each process characteristic item are recorded; the field name of the process characteristic output by different detection equipment is uniformly mapped to the internal standard field specification, the original field name and mapping relationship are retained, and the standard field dictionary table is established; the process characteristic field with composite structure is flattened according to the unified rule, which is split into a one-dimensional process characteristic field set, and the field and production batch number association index table is established; the physical unit representation of all collected fields is unified, and the international unit system is used for conversion, and the unit conversion success rate, unit abnormal field proportion and unit standardization completion rate are recorded; the collected process characteristic data, production environment data and processing process data are normalized to eliminate the dimensional and scale differences of process parameters between batches; the total number of sampling parameters, the number of historical production batches and the actual processing value of each process parameter, the rule correction value, the number of abnormal monitoring indicators, the characteristic stable value and the batch characteristic standardization completion rate of each batch product are counted, and the standardized and normalized process characteristic data, production environment data and processing process data are stored to construct the standardized data set.
[0013] Further, the non-linear coupling relationship between different process parameters is comprehensively evaluated based on the standardized data set and the process parameter characteristics of the historical batches, and the process parameter interaction value for the current batch is generated. The specific steps are as follows: the total number of collected process parameters and the real-time collection value and historical collection average value of each process parameter of the current batch are extracted from the standardized database, the difference between the real-time collection value and the historical collection average value is calculated and the absolute value is taken; the instantaneous fluctuation amplitude of each process parameter of the current batch at the current time is extracted, the difference absolute value between the real-time collection value and the historical collection average value is divided by the instantaneous fluctuation amplitude to obtain the deviation degree value; the deviation degree values of all collected process parameters are multiplied item by item to obtain the overall process deviation product; the change rate of each process parameter at the current time is extracted and added to the stability factor to obtain the change stability value, and the change stability values of all collected process parameters are summed to obtain the overall change trend value; the overall process deviation product is divided by the overall change trend value, and the logarithm is taken to obtain the process parameter interaction value of the current batch in the production process.
[0014] Further, the specific steps of the process parameter interaction value-based optimization of subsequent data acquisition frequency and dynamic adjustment strategy are: real-time comparison of the process parameter interaction value with the interaction stability threshold, which includes a first stability threshold and a second stability threshold; when the process parameter interaction value is greater than or equal to the first stability threshold, it is determined to be a high stability section, the current data acquisition and processing strategy is kept unchanged, and process optimization derivation is directly based on the evaluation value without additional introduction of numerical compensation terms; when the process parameter interaction value is greater than or equal to the second stability threshold and less than the first stability threshold, it is determined to be a medium stability section, a numerical warning state is entered, a moderate compensation strategy is executed, including dynamic adjustment of data acquisition frequency, expansion of the sampling window range, and introduction of a small positive number into the process parameter change rate and fluctuation amplitude, enhancing the numerical stability of the formula, while starting formula operation error monitoring to ensure the continuity and reliability of the calculation process; when the process parameter interaction value is less than the second stability threshold, it is determined to be a low stability section, the process optimization derivation based on the current data is immediately suspended, a numerical abnormality emergency correction mechanism is started, historical batch process characteristics are re-extracted, multi-batch dynamic weighted average parameters are constructed to replace the current batch abnormal parameters for formula calculation, and the current batch is listed as a high-priority data quality traceability object to record data deviation characteristics and stability failure conditions.
[0015] Further, the specific steps of the process parameter interaction value-based optimization of subsequent data acquisition frequency and dynamic adjustment strategy are: real-time comparison of the process parameter interaction value with the interaction stability threshold, which includes a first stability threshold and a second stability threshold; when the process parameter interaction value is greater than or equal to the first stability threshold, it is determined to be a high stability section, the current data acquisition and processing strategy is kept unchanged, and process optimization derivation is directly based on the evaluation value without additional introduction of numerical compensation terms; when the process parameter interaction value is greater than or equal to the second stability threshold and less than the first stability threshold, it is determined to be a medium stability section, a numerical warning state is entered, a moderate compensation strategy is executed, including dynamic adjustment of data acquisition frequency, expansion of the sampling window range, and introduction of a small positive number into the process parameter change rate and fluctuation amplitude, enhancing the numerical stability of the formula, while starting formula operation error monitoring to ensure the continuity and reliability of the calculation process; when the process parameter interaction value is less than the second stability threshold, it is determined to be a low stability section, the process optimization derivation based on the current data is immediately suspended, a numerical abnormality emergency correction mechanism is started, historical batch process characteristics are re-extracted, multi-batch dynamic weighted average parameters are constructed to replace the current batch abnormal parameters for formula calculation, and the current batch is listed as a high-priority data quality traceability object to record data deviation characteristics and stability failure conditions.
[0016] Further, the adjustment range interval and dynamic correction strategy of each key process parameter are further updated according to the process sensitivity analysis result, and the specific steps are as follows: comparing the process sensitivity evaluation value with the sensitivity threshold value in real time, the sensitivity threshold value includes a first sensitivity threshold value and a second sensitivity threshold value; when the process sensitivity evaluation value is greater than or equal to the first sensitivity threshold value, it is determined that it is in a high sensitivity section, the existing process parameter configuration and data acquisition frequency are kept unchanged, the production process is continuously executed, and a high-priority data archiving mode is entered, and the energy consumption index, raw material consumption curve and operation log are recorded in real time; when the process sensitivity evaluation value is greater than or equal to the second sensitivity threshold value and less than the first sensitivity threshold value, it is determined that it is in a medium sensitivity section, a parameter fine-tuning mechanism is triggered, the reaction temperature, feed flow and catalyst addition ratio are dynamically adjusted according to the real-time monitoring data trend, the sampling frequency is simultaneously increased, the energy consumption and raw material use data acquisition cycle is shortened, the monitoring density on energy efficiency sensitive parameters and resource consumption key indicators is enhanced, a medium energy efficiency risk prompting function is activated, the operator is reminded to pay attention to process fluctuation changes to prevent energy efficiency abnormalities from further expanding; when the process sensitivity evaluation value is less than the second sensitivity threshold value, it is determined that it is in a low sensitivity section, the production operation based on the current parameters is immediately suspended, an emergency parameter correction mechanism is started, and the current batch is listed as a high-priority energy efficiency abnormality traceability and risk assessment object, the energy consumption deviation characteristics, resource waste trend and abnormal operation log are comprehensively recorded.
[0017] Further, the comprehensive dynamic matching degree of the current batch under the nonlinear process characteristics and the strong coupling parameter characteristics is analyzed by taking the current batch process parameter interaction value and the process sensitivity analysis result as inputs, and the specific steps are as follows: obtaining the process parameter interaction value and the process sensitivity evaluation value, taking the process sensitivity evaluation value to the power of the sensitivity adjustment factor, multiplying the process parameter interaction value, obtaining the process interaction sensitivity combination value; extracting the collection data integrity rate of the current process parameter, multiplying the trend sensitivity coefficient and the second-order inverse absolute value of the collection data integrity rate, and summing the nonlinear adjustment index power of the first-order derivative absolute value of the collection data integrity rate, obtaining the collection data sensitivity value; taking the logarithm of the process interaction sensitivity combination value divided by the collection data sensitivity value plus one, obtaining the dynamic optimization adjustment value of the current process parameter.
[0018] Further, the step of generating the optimal process adjustment optimization suggestion based on the comprehensive dynamic matching degree analysis result is: real-time comparison of the dynamic optimization adjustment value and the process control response threshold, the process control response threshold includes a first control threshold and a second control threshold: when the dynamic optimization adjustment value is greater than or equal to the first control threshold, it is determined as a high matching adaptation section, the current process parameter setting is kept unchanged, and an optimization suggestion for maintaining process stability is generated, the key parameter items for subsequent monitoring and preventive adjustment instructions are specified, the operator is assisted to continuously monitor the process fluctuation trend, and it is ensured that the production process is efficiently operated in the stable interval; when the dynamic optimization adjustment value is greater than or equal to the second control threshold and less than the first control threshold, it is determined as a medium matching adaptation section, and a process dynamic fine adjustment state is entered, a local parameter self-adaptive adjustment mechanism is started according to the real-time collected process characteristic data change trend, the key process parameter settings of reaction temperature, feed flow and catalyst ratio are dynamically refined, the matching degree of batch characteristics and process parameters is improved, and process optimization suggestions are generated based on the characteristic vector change trend, the control range and adjustment step of the related parameters are adjusted by the operator, the process setting is dynamically corrected to adapt to the batch characteristic change, and it is ensured that the process parameters remain in the optimal response state in the multi-dimensional characteristic space; when the dynamic optimization adjustment value is less than the second control threshold, it is determined as a low matching adaptation section, the process execution process of the current batch is immediately suspended, an emergency process rollback mechanism is started, and a predefined emergency process parameter template is quickly switched to ensure production safety; at the same time, based on the current batch characteristic characteristics, an optimization adjustment instruction for the abnormal batch is output, a process parameter rollback path, a key characteristic correction direction and a next step dynamic optimization strategy suggestion are clearly proposed, and the operator is helped to restore the process stability in the shortest time, the real-time adaptability and process control robustness of the system to the nonlinear complex characteristic batch are improved, and a closed-loop dynamic optimization response mechanism is constructed.
[0019] The second aspect of the application provides a dimethyl sulfoxide production optimization system based on a deep neural network, comprising: a data acquisition and preprocessing module, a parameter relationship analysis module, a sensitivity analysis module, and a dynamic optimization module, wherein: the data acquisition and preprocessing module is configured to collect process characteristic data, production environment data, and processing process data of the current batch by deploying multiple types of detection terminals, and to perform standardization and normalization preprocessing on the collected process characteristic data, production environment data, and processing process data to construct a standardized data set; the parameter relationship analysis module is configured to evaluate the nonlinear coupling relationship between different process parameters based on the standardized data set and historical batch process parameter characteristics, generate process parameter interaction values for the current batch, and optimize the subsequent data acquisition frequency and the priority of the dynamic adjustment strategy according to the process parameter interaction values; the sensitivity analysis module is configured to perform process sensitivity analysis on each process parameter in the dimethyl sulfoxide production process under the characteristics of the current batch based on the standardized data set, and then update the adjustment range interval and dynamic correction strategy of each key process parameter according to the process sensitivity analysis results; and the dynamic optimization module is configured to take the process parameter interaction values and the process sensitivity analysis results of the current batch as inputs, comprehensively analyze the comprehensive dynamic matching degree of the current batch under nonlinear process characteristics and strongly coupled parameter characteristics, and generate optimal process adjustment optimization suggestions based on the comprehensive dynamic matching degree analysis results.
[0020] Advantages
[0021] The application has the following advantages:
[0022] (1) The dimethyl sulfoxide production optimization method and system based on a deep neural network dynamically adjusts process parameter settings based on batch process characteristics recognition and real-time monitoring feedback, thereby realizing personalized matching and stable execution of the production process, and effectively solving the problems of fixed process parameters, insufficient adaptability, and large product yield fluctuations in the prior art.
[0023] (2) The dimethyl sulfoxide production optimization method and system based on a deep neural network introduces a comprehensive optimization mechanism based on joint calculation of process interaction characteristic evaluation values, process sensitivity evaluation values, and trend change factors, thereby improving the real-time matching degree and control sensitivity in the process execution, and effectively solving the problems of parameter optimization relying on static rules and response lag in the prior art.
[0024] (3) The dimethyl sulfoxide production optimization method and system based on a deep neural network constructs a batch characteristic vector modeling system based on standardization and normalization processing of multi-source process monitoring data, integrates historical production data to realize dynamic recommendation and adaptive update of process parameters, thereby enhancing the adaptability of the system to batch characteristic differences, and effectively solving the problems of single batch process parameter configuration and lack of targeting in the prior art.
[0025] (4), the dimethyl sulfoxide production optimization method and system based on deep neural network, through unified multi-source monitoring data acquisition standard, structured preprocessing and process parameter dynamic archiving mechanism, and then improve the fusion integrity of production process data and the expansion efficiency of process history knowledge base, effectively solve the problem of non-standard data collection and difficult process experience accumulation in the prior art.
[0026] Of course, implementing any product of the present application does not necessarily require all the advantages described above to be achieved at the same time. BRIEF DESCRIPTION OF DRAWINGS
[0027] Fig. 1 For the dimethyl sulfoxide production optimization method based on deep neural network of the present application flow chart;
[0028] Fig. 2 For the dimethyl sulfoxide production optimization system structure diagram based on deep neural network of the present application;
[0029] Fig. 3 For the process sensitivity evaluation value column chart involved in the present application. DETAILED DESCRIPTION
[0030] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.
[0031] Please refer to Figs. 1-3The embodiment of the present application provides a technical scheme: a dimethyl sulfoxide production optimization method and system based on a deep neural network, comprising S1, deploying multiple types of detection terminals to collect process characteristic data, production environment data and processing process data of the current batch, and performing standardization and normalization preprocessing on the collected process characteristic data, production environment data and processing process data to construct a standardized data set; S2, based on the standardized data set, combining the historical batch process parameter characteristics, comprehensively evaluating the nonlinear coupling relationship between different process parameters, generating process parameter interaction values for the current batch, and optimizing the subsequent data collection frequency and the priority of the dynamic adjustment strategy according to the process parameter interaction values; S3, based on the standardized data set, performing process sensitivity analysis on each process parameter in the dimethyl sulfoxide production process under the characteristics of the current batch, and then updating the adjustment amplitude interval and the dynamic correction strategy of each key process parameter according to the process sensitivity analysis result; S4, taking the process parameter interaction value and the process sensitivity analysis result of the current batch as input, comprehensively analyzing the comprehensive dynamic matching degree of the current batch under the nonlinear process characteristics and the strong coupling parameter characteristics, and generating an optimal process adjustment optimization suggestion based on the comprehensive dynamic matching degree analysis result.
[0032] Specifically, the deployment of multiple types of detection terminals collects the current batch of process characteristic data, production environment data and processing process data. The specific steps are as follows: collecting process characteristic data including raw material ratio, reaction temperature curve, reaction pressure curve, catalyst concentration, stirring rate change, raw material flow fluctuation, reaction time, product yield, product purity and energy consumption index through production process detection equipment, and synchronously collecting key reaction process monitoring data including byproduct content change trend, reaction intermediate concentration distribution and reaction heat release rate; and recording each process characteristic data in detail, including measurement unit, measurement method, detection accuracy, unit energy consumption change, detection equipment model, data collection time interval, ratio of historical measurement mean value and standard deviation, deviation degree of current process characteristic from historical mean value, instantaneous fluctuation amplitude and sampling time environmental temperature, humidity and air pressure conditions, to ensure the comprehensiveness and timeliness of data collection, and to provide accurate basic data support for subsequent characteristic modeling and process optimization; collecting production environment data including reaction kettle surrounding temperature, pressure, humidity, gas flow rate, heat exchange efficiency, equipment running time, raw material consumption rate and condensation efficiency through process environment monitoring equipment, and extending to collect auxiliary process environment characteristics including equipment surface temperature rise rate, condensation system return flow fluctuation and energy consumption load change curve; and recording each process environment data in detail, including collection frequency, process operation frequency, change rate, real-time collection value, resource consumption number, monitoring period, historical monthly mean value and standard deviation, data missing rate, abnormal fluctuation identification and current process section number and production batch number, to improve the granularity and dynamic response capability of environment characteristic monitoring; at the same time, recording processing process data including total number of collected process parameters, time stamp of each collected data, sampling interval, measurement time length, data completeness rate, historical collection mean value, unit energy consumption change mean value, data distribution uniformity index, batch data coverage rate and environment monitoring synchronization rate, establishing multi-dimensional and multi-batch data link, supporting subsequent process parameter normalization processing, characteristic vector construction and dynamic optimization analysis, and improving the process perception accuracy and intelligent control capability of the overall production process.
[0033] In the embodiment, by systematically collecting and recording process characteristic data, production environment data and processing process data involved in the production process of dimethyl sulfoxide, including reactant material ratio, temperature and pressure change trend, catalyst concentration fluctuation, energy consumption load curve and reaction byproduct dynamic characteristics, supplemented by real-time environment monitoring and operation parameter frequency recording, a complete and fine multi-source process data chain is constructed; by synchronously recording the collection accuracy, fluctuation amplitude, change rate and resource consumption index of each data, the timeliness and accuracy of the data are enhanced. This step provides a high-quality data basis for subsequent process characteristic standardization processing, characteristic vector modeling and deep neural network optimization analysis, significantly improves the perception accuracy, dynamic response capability and intelligent control level of the process, and supports the evolution of dimethyl sulfoxide production process towards precision, dynamic and self-adaptive direction.
[0034] Specifically, the collected process characteristic data, production environment data and processing process data are standardized and normalized for preprocessing, and the specific steps for constructing the standardized data set are as follows: through a unified time synchronization mechanism, the time stamp of all process characteristic data and production environment data is standardized to ensure that different data sources are synchronized and consistent under the same time reference, and the time alignment success rate, time drift average and maximum time drift value of each production batch data are recorded to improve the timeliness and accuracy of data fusion; the null values, invalid extreme values, format abnormal fields and measurement error out-of-limit values in the process characteristic data, production environment data and processing process data are cleaned, and further repeated records, logical conflict fields and cross-batch interference data are removed, and the missing rate, abnormal proportion, removal ratio and effective data retention rate after cleaning of each process characteristic item are recorded to improve the data quality stability;
[0035] The field names of process characteristics output by different detection equipment are uniformly mapped to internal standard field specifications to construct a unified rule system for data fields across devices and batches, the original field names and mapping relationships are retained, a standard field dictionary table is established, and the field name mapping success rate and field conflict resolution ratio are recorded to ensure data consistency and traceability; process characteristic fields with complex structures are flattened according to unified rules and split into a one-dimensional process characteristic field set, the field splitting rules and splitting levels are labeled, and an associated index table of fields and production batch numbers is established to improve the processing efficiency and compatibility of complex process characteristics in subsequent modeling processes;
[0036] The physical unit representation of all collected fields is unified and converted using the International System of Units, taking into account unit correction and anomaly detection of historical batch non-standard data, and recording the unit conversion success rate, unit abnormal field proportion, unit standardization completion rate and unit anomaly correction log to ensure dimensional consistency and data comparability; the collected process characteristic data, production environment data and processing process data are normalized, the maximum and minimum value normalization method is used according to the numerical distribution interval, historical mean and standard deviation characteristics of each field to eliminate the dimensional and scale differences of process parameters between batches, and to improve the uniformity of data distribution and the convergence speed of deep modeling;
[0037] The total number of sampling parameters, the number of historical production batches, and the actual processing values of each process parameter, the rule correction values, the number of abnormal monitoring indicators, the characteristic stable values and the batch characteristic standardization completion rate of each batch of products are counted to evaluate the integrity and reliability of the overall data set; the standardized and normalized process characteristic data, production environment data and processing process data are stored and version-controlled batch by batch to construct a dynamically updated standardized data set, providing high-quality data foundation support for subsequent deep neural network modeling and real-time process optimization.
[0038] In this embodiment, the time reference alignment of multi-source process characteristic data and production environment data is achieved through a unified time synchronization mechanism, improving the synchronization and timeliness of data fusion; the data quality and stability are significantly improved through null value elimination, abnormality detection and format standardization processing; the consistency and comparability of data between different sources and batches are ensured through unified field specification, composite field splitting and physical unit standardization; the dimensional and scale differences between different process parameters are eliminated through normalization processing, enhancing the uniformity of data distribution and the adaptability of modeling; a standardized data set with high integrity and high reliability is constructed through dynamic statistics and version management. The standardization and normalization processing steps provide a unified, standardized and high-quality data basis for subsequent characteristic vector modeling, process state evaluation and adaptive optimization based on deep neural networks, significantly improving the intelligentization, precision and dynamic response capability of dimethyl sulfoxide production process control.
[0039] Specifically, based on the standardized data set and combining the process parameter characteristics of each batch, the nonlinear coupling relationship between different process parameters is comprehensively evaluated, and the process parameter interaction value for the current batch is generated. The specific steps are as follows: the total number of collected process parameters and the real-time collection value and historical collection mean value of each process parameter of the current batch are extracted from the standardized database, the difference between the real-time collection value and the historical collection mean value is calculated and the absolute value is taken; the instantaneous fluctuation amplitude of each process parameter of the current batch is extracted, and the absolute value of the difference between the real-time collection value and the historical collection mean value is divided by the instantaneous fluctuation amplitude to obtain the deviation degree value; the deviation degree values of all collected process parameters are multiplied item by item to obtain the overall process deviation product; the change rate of each process parameter at the current time is extracted and added to the stability factor to obtain the change stability value, and the change stability values of all collected process parameters are summed to obtain the overall change trend value; the overall process deviation product is divided by the overall change trend value and the logarithm is taken to obtain the process parameter interaction value of the current batch in the production process.
[0040] The process parameter interaction value calculation formula is:
[0041]
[0042] In the formula, n represents the total number of collected process parameters, which is used to limit the size of the process parameter set in dynamic analysis, supporting multi-parameter interaction characteristic fusion calculation; X k (t) represents the real-time collection value of the kth process parameter, which is used to represent the actual measurement state of each process parameter in the production process of the current batch; represents the historical collection mean value of the kth process parameter, which is used as a process parameter reference benchmark to measure the deviation degree of the real-time collection value from the historical level; Y k(t) represents the instantaneous fluctuation amplitude of the kth process parameter at time t, which is used to measure the short-period fluctuation of the process parameter, and reflects the stability characteristics of the process execution process; Z k (t) represents the change rate of the kth process parameter at time t, which is used to characterize the dynamic trend of the process parameter with time, and supports the dynamic response capability evaluation; λ represents a stability factor, which is in the range of 0.1 to 0.5, and is dynamically set by comprehensively analyzing the fluctuation amplitude and change rate of each process parameter in the collection period. When the fluctuation amplitude of the process parameter is small, the change trend is stable, and the data continuity is good, the stability factor is correspondingly increased to enhance the weight contribution of the stable parameter in the process interaction characteristic calculation, and to ensure the reliability support of the low fluctuation and high stability process parameter to the overall process state evaluation. If the process parameter has high-frequency mutation, large fluctuation amplitude or intermittent collected data, the stability factor is appropriately reduced to suppress the interference of abnormal fluctuation on the process interaction characteristic evaluation result, and to improve the robustness and dynamic response accuracy of the interaction characteristic calculation.
[0043] In the embodiment, by fusing the real-time deviation amplitude, instantaneous fluctuation intensity and change rate information of the process parameters, a multi-dimensional and dynamic process parameter interaction strength evaluation index is established. The formula is based on the deviation normalization processing and change trend correction of the process parameters, balances the differences in data fluctuation and stability of different batches and different parameters, and improves the numerical stability and dynamic response capability of the interaction characteristic evaluation. The interaction characteristic evaluation value provides basic quantitative support for subsequent process characteristic modeling, dynamic process optimization and abnormal state recognition, significantly enhances the real-time perception, precise control and intelligent optimization level of the dimethyl sulfoxide production process.
[0044] Specifically, and according to the process parameter interaction value, the subsequent data collection frequency and the priority of the dynamic adjustment strategy are optimized. The specific steps are as follows: real-time comparison of the process parameter interaction value and the interaction stability threshold, the interaction stability threshold includes a first stability threshold and a second stability threshold, which is used to determine the stability level and the corresponding coping strategy in the dynamic interaction process of the process parameter:
[0045] When the process parameter interaction value is greater than or equal to the first stability threshold, it is determined to be a high stability section, indicating that the interaction coupling relationship between the current process parameters is stable, the fluctuation amplitude is controlled, and the change trend is smooth, and the current data acquisition and processing strategy is maintained unchanged, the original sampling frequency, data cleaning rule and standardization parameter setting are continued to be used, and process optimization derivation is directly based on the evaluation value, without additional introduction of numerical compensation term, ensuring that the process optimization process quickly converges under stable data support, improving decision efficiency and optimization reliability; when the process parameter interaction value is greater than or equal to the second stability threshold and less than the first stability threshold, it is determined to be a medium stability section, indicating that there is a certain degree of interaction fluctuation and trend drift between the process parameters, entering the numerical warning state, starting the moderate compensation strategy, dynamically adjusting the data acquisition frequency, expanding the sampling window range to enhance the sampling density and data continuity, and introducing a small positive correction term in the change rate and fluctuation amplitude of each process parameter, improving the numerical stability and convergence characteristics of the formula calculation, and simultaneously starting the formula operation error monitoring program, tracking the error accumulation trend and calculation accuracy change in real time, ensuring the continuity of the calculation process and the reliability of the final result, and reducing the optimization deviation risk caused by data disturbance; when the process parameter interaction value is less than the second stability threshold, it is determined to be a low stability section, indicating that the current data has severe fluctuations, abnormal jumps or trend reversals, which is difficult to support effective process optimization decision, immediately suspending the process optimization derivation based on the current data, activating the numerical abnormal emergency correction mechanism, re-extracting the historical batch data similar to the current batch process characteristics, constructing a multi-batch dynamically weighted average parameter set based on the weighted characteristic similarity, replacing the current batch abnormal parameters to participate in formula calculation, avoiding abnormal data interference optimization decision, and at the same time listing the current batch as a high-priority data quality traceability object, recording the data deviation characteristics, interaction abnormal trend and stability failure, for subsequent abnormal batch analysis and data quality improvement, improving the robustness and adaptability of low stability process data.
[0046] In this embodiment, by comparing the process parameter interaction value with the interaction stability threshold in real time, the stability level of the current production data is dynamically determined, and the data processing and process optimization strategy is adaptively adjusted based on different stability segments. In the high stability segment, the original data acquisition and processing flow is maintained to ensure the rapid response and stable convergence of the process optimization decision; in the medium stability segment, the sampling frequency adjustment, window expansion and numerical fine-tuning correction are performed to enhance the data continuity and calculation stability and reduce the interference of data fluctuation on optimization derivation; in the low stability segment, the process optimization based on the current batch data is suspended, and the historical multi-batch weighted replacement strategy is adopted to avoid the influence of abnormal data on the optimization result, while realizing high-priority tracing and data quality monitoring of abnormal batches. Through this step, the robustness and adaptability of the process optimization process under different data stability conditions are effectively improved, ensuring the accuracy, reliability and dynamic response capability of the optimization decision in the dimethyl sulfoxide production process.
[0047] Specifically, based on the standardized data set, the process sensitivity of each process parameter in the dimethyl sulfoxide production process under the current batch characteristics is analyzed as follows: the instantaneous fluctuation amplitude of each process parameter in the current batch, the caused unit energy consumption change and the average unit energy consumption change of the historical batch are extracted from the standardized database, the absolute value of the difference between the unit energy consumption change in the current batch and the average unit energy consumption change of the historical batch is calculated, and the instantaneous fluctuation amplitude is divided by the instantaneous fluctuation amplitude to obtain the energy consumption normalized value of each process parameter; the process operation frequency and resource consumption of each process parameter in the current batch are extracted, and the process operation frequency is divided by the resource consumption coefficient to obtain the resource utilization efficiency value of each process parameter; the energy consumption normalized value and the resource utilization efficiency value are multiplied and then added by one, and the logarithm is taken to obtain the process sensitivity evaluation value of the current process parameter.
[0048] The process sensitivity evaluation value calculation formula is:
[0049]
[0050] In the formula, E k represents the unit energy consumption change caused by the kth process parameter in the current batch, which is used to represent the actual influence degree of the current process parameter on the energy consumption index and reflects the sensitivity of process control to production energy efficiency; E a represents the average unit energy consumption change of the historical batch, which is used as a reference benchmark to measure the deviation degree of the current process parameter energy consumption change relative to the historical level; Y k (t) represents the instantaneous fluctuation amplitude of the kth process parameter in the current production cycle, which is used to quantify the fluctuation intensity of the process parameter in the short cycle and reflects the stability of the process execution process; G krepresents the kth process parameter in the current batch of process operation frequency, used to characterize the operation of the process parameter active degree, reflecting its dynamic regulation intensity in the process of execution; R k represents the resource consumption number of the kth process parameter, used to quantify the corresponding resource use intensity of the process parameter in the production process, reflecting the resource utilization efficiency of the process control process.
[0051] In this embodiment, the current batch unit energy consumption change of process parameter P1 is set to 120.5, the unit energy consumption change average is set to 110.0, the instantaneous fluctuation amplitude is set to 5.2, the process operation frequency is set to 30, and the resource consumption coefficient is set to 45; the current batch unit energy consumption change of process parameter P2 is set to 95.0, the unit energy consumption change average is set to 90.5, the instantaneous fluctuation amplitude is set to 4.5, the process operation frequency is set to 25, and the resource consumption coefficient is set to 40; the current batch unit energy consumption change of process parameter P3 is set to 150.0, the unit energy consumption change average is set to 140.0, the instantaneous fluctuation amplitude is set to 8.0, the process operation frequency is set to 20, and the resource consumption coefficient is set to 50; the current batch unit energy consumption change of process parameter P4 is set to 180.2, the unit energy consumption change average is set to 170.0, the instantaneous fluctuation amplitude is set to 10.5, the process operation frequency is set to 35, and the resource consumption coefficient is set to 55; the current batch unit energy consumption change of process parameter P5 is set to 130.5, the unit energy consumption change average is set to 128.0, the instantaneous fluctuation amplitude is set to 6.2, the process operation frequency is set to 40, and the resource consumption coefficient is set to 48. The process sensitivity evaluation value of each type of process parameter is calculated, as shown in Table 1 Process Sensitivity Evaluation Value Data Table.
[0052] Table 1 Process Sensitivity Evaluation Value Data Table
[0053]
[0054] As shown in Fig. 3 , the process sensitivity evaluation value histogram provided by the present application example. As shown in Table 1 and Fig. 3It can be seen that the process sensitivity evaluation value of process parameter P1 is the highest, indicating that the unit energy consumption change amount deviates greatly, the fluctuation intensity is low, and the process operation frequency is high, which comprehensively reflects that the energy efficiency sensitivity of the process parameter in the current batch production process is high, the regulation and response are timely, and the parameter can be preferentially adjusted and optimized to improve the energy efficiency stability and regulation accuracy of the overall production process; the process sensitivity evaluation value of process parameter P4 is the lowest, although the operation frequency is moderate, the unit energy consumption change is small and the fluctuation range is large, resulting in a low overall sensitivity level, which will automatically reduce the priority of dynamic regulation of the parameter, retain it as a secondary optimization object, reduce resource occupation and invalid scheduling, and improve the overall process optimization efficiency. The process sensitivity evaluation value column chart directly reflects the energy efficiency sensitivity distribution characteristics of each process parameter in the current production batch, the higher the evaluation value, the more sensitive the process parameter, and the more inclined to preferentially adjust the parameter to achieve precise regulation and energy efficiency optimization.
[0055] Specifically, the adjustment range interval and dynamic correction strategy of each key process parameter are further updated according to the process sensitivity analysis results, and the specific steps are as follows: real-time comparison of process sensitivity evaluation value and sensitivity threshold, the sensitivity threshold includes first sensitivity threshold and second sensitivity threshold:
[0056] When the process sensitivity evaluation value is greater than or equal to the first sensitivity threshold, it is determined to be a high sensitivity section, indicating that the current process parameter has a high sensitivity to production energy efficiency changes, the existing process parameter configuration and data acquisition frequency are kept unchanged, the established production process is continuously executed to avoid system fluctuations caused by frequent adjustments, and a high-priority data archiving mode is entered to record energy consumption indicators, raw material consumption curves and operation logs in real time, expand the record of catalyst usage, stirring rate change trend and reaction heat change characteristics, and ensure the complete data traceability of high-sensitivity batch production process, providing high-quality historical data support for subsequent process optimization and energy efficiency model updating;
[0057] When the process sensitivity evaluation value is greater than or equal to the second sensitivity threshold and less than the first sensitivity threshold, it is determined to be a medium sensitivity section, indicating that the process parameter has a certain degree of sensitivity to energy efficiency changes but the fluctuation is controllable, the parameter fine-tuning mechanism is automatically triggered, the key process setting parameters such as reaction temperature, feed flow and catalyst addition ratio are dynamically adjusted according to the real-time monitoring data trend, the energy efficiency output performance is optimized, the sampling frequency is simultaneously improved, the energy consumption and raw material usage data acquisition period is shortened, the monitoring density on energy sensitive parameters and resource consumption key indicators is enhanced, the device load change, exhaust rate and intermediate product concentration change information are supplemented, the medium energy efficiency risk prompt function is activated, the risk warning report is pushed in real time, reminding the operator to pay attention to the process fluctuation trend, preventing the energy efficiency from further expanding, and ensuring the controllability and energy efficiency stability of the process production process;
[0058] When the process sensitivity evaluation value is less than the second sensitivity threshold, it is determined that it is a low sensitivity section, indicating that the current process parameter responds slowly to energy efficiency changes or there is an abnormal deviation, the production operation based on the current parameter is immediately suspended, the emergency parameter correction mechanism is started, the historical optimal process parameter template is automatically rolled back to quickly restore the process running stability, and the current batch is listed as a high-priority energy efficiency abnormality traceability and risk assessment object, the energy consumption deviation characteristics, resource waste trend, abnormal operation log and production environment change data are expanded to form a complete abnormal batch characteristic archive, supporting subsequent in-depth abnormality analysis and process optimization strategy adjustment, and improving the overall energy efficiency management level and abnormal risk prevention and control capability of the system.
[0059] In the embodiment, by comparing the process sensitivity evaluation value with the sensitivity threshold in real time, different sensitivity sections are dynamically identified, and the process parameter configuration and data acquisition strategy are intelligently adjusted based on the sensitivity level. For the high sensitivity section, the process parameter setting and data acquisition frequency remain unchanged, the archiving of key energy efficiency indicators and process characteristic data is strengthened to ensure that high-quality data support process optimization and model updating; for the medium sensitivity section, the process parameter fine-tuning mechanism and data acquisition density improvement measures are triggered to enhance the dynamic perception and risk warning capability of the energy efficiency change trend; for the low sensitivity section, the abnormal batch production is suspended, the historical optimal process parameter template is rolled back, the abnormal batch traceability and comprehensive characteristic recording are performed, and the energy efficiency abnormality prevention and control and process robustness are improved. Through this step, the dynamic response speed, energy efficiency management level and overall optimization capability of the process control in the dimethyl sulfoxide production process are effectively improved, supporting the intelligent, precise and efficient upgrading of the production process.
[0060] Specifically, the current batch process parameter interaction value and the process sensitivity analysis result are input, and the comprehensive dynamic matching degree of the current batch under the nonlinear process characteristic and the strong coupling parameter characteristic is analyzed. The specific steps are: obtaining the process parameter interaction value and the process sensitivity evaluation value, multiplying the process sensitivity evaluation value by the sensitivity adjustment factor power, and multiplying the process parameter interaction value to obtain the process interaction sensitivity combination value; extracting the collection data integrity rate of the current process parameter, multiplying the trend sensitivity coefficient by the second-order inverse absolute value of the collection data integrity rate, and summing the nonlinear adjustment index power of the first-order derivative absolute value of the collection data integrity rate to obtain the collection data sensitivity value; taking the logarithm of the process interaction sensitivity combination value divided by the collection data sensitivity value plus one to obtain the dynamic optimization adjustment value of the current process parameter.
[0061] The dynamic optimization adjustment value calculation formula is:
[0062]
[0063] In the formula, R(t) represents the process parameter interaction value, which is used to comprehensively quantify the interaction coupling characteristics of each process parameter in the dynamic change process, and reflects the nonlinear coupling strength in the process execution process; S k represents the process sensitivity evaluation value, which is used to measure the response strength of each process parameter to the energy efficiency change and resource consumption, and reflects the parameter sensitivity level in the process regulation process; P k (t) represents the collection data integrity rate of the kth process parameter, which is used to characterize the timeliness and integrity of the process parameter sampling data, and reflects the data basis quality; θ k represents the trend sensitivity coefficient of the kth process parameter, which is in the range of 0.1 to 0.5, and is dynamically set by comprehensively analyzing the trend fluctuation frequency and trend continuity of the process parameter in the production process; when the process parameter change trend is stable and the trend continuity is good, the trend sensitivity coefficient is correspondingly increased to enhance the contribution weight of the parameter in the dynamic trend analysis, ensure the accuracy of the trend change stable parameter in capturing the process fluctuation trend, and if the process parameter change trend is severe or there is a trend breakpoint, the trend sensitivity coefficient is appropriately reduced to weaken the interference of the trend mutation data on the overall trend analysis accuracy, and improve the robustness and reliability of the process trend perception; β represents the sensitivity adjustment factor, which is in the range of 0.2 to 0.6, and is dynamically set by comprehensively analyzing the fluctuation amplitude of the sensitivity evaluation value of the process parameter and the historical sensitivity distribution characteristics; when the process parameter sensitivity is stable, the fluctuation amplitude is small, and the historical sensitivity level is stable, the sensitivity adjustment factor is correspondingly increased to enhance the role weight of the sensitivity evaluation value in the process dynamic regulation, improve the priority identification ability of the high sensitivity parameter in the optimization decision, and if the sensitivity evaluation value fluctuates greatly, the sensitivity adjustment factor is appropriately reduced to avoid the interference of abnormal sensitivity fluctuation on the process optimization process, and improve the stability and robustness of the process regulation decision; δ represents the nonlinear adjustment index factor, which is in the range of 0.1 to 0.7, and is dynamically set by comprehensively analyzing the dynamic stability and change continuity of the process parameter data change rate; when the process parameter data change rate is stable and the change continuity is good, the nonlinear index adjustment factor is correspondingly increased to enhance the response contribution of the stable change parameter in the dynamic regulation formula, and improve the process change trend capture accuracy; if the process parameter data change rate fluctuates severely, the nonlinear index adjustment factor is appropriately reduced to weaken the interference of severe change on the overall dynamic evaluation result, and enhance the robustness and abnormal tolerance ability of the formula operation.
[0064] In this embodiment, by combining the interactive characteristic evaluation value, sensitivity evaluation value, data change trend and change rate of process parameters, a multi-dimensional dynamic optimization evaluation index is constructed. The formula introduces sensitivity adjustment factor and nonlinear index adjustment factor to dynamically adjust the weight proportion of sensitivity evaluation value and data change rate in the overall evaluation, and corrects the response characteristics of the trend sensitivity coefficient to the parameter change trend, which improves the adaptive identification and response ability of different fluctuation characteristics of process parameters. The comprehensive evaluation value provides a basis for subsequent process parameter priority ranking, dynamic optimization strategy derivation and abnormal batch rapid screening, effectively enhances the intelligent, precise and dynamic optimization ability of the dimethyl sulfoxide production process, and improves the overall energy efficiency management level and the stability and robustness of the production process.
[0065] Specifically, the specific steps of generating the optimal process adjustment optimization suggestion based on the comprehensive dynamic matching degree analysis result are: real-time comparison of the dynamic optimization adjustment value and the process control response threshold, and the process control response threshold includes the first control threshold and the second control threshold: when the dynamic optimization adjustment value is greater than or equal to the first control threshold, it is determined as a high matching adaptation section, indicating that the current process parameter combination is highly matched with the batch characteristics, the process execution process is stable, the production energy efficiency is at an excellent level, the system keeps the current process parameter setting unchanged, avoids invalid adjustment to cause fluctuations, and automatically generates optimization suggestions to maintain process stability, clearly indicates the subsequent monitoring key parameter items, trend drift monitoring interval and preventive adjustment guidelines, assisting the operator to continuously monitor the process fluctuation trend and key energy efficiency indicators, dynamically warning potential deviation risks, ensuring the production process to run efficiently within the stable interval, and improving the production continuity and energy efficiency consistency;
[0066] When the dynamic optimization adjustment value is greater than or equal to the second control threshold and less than the first control threshold, it is determined as a medium matching adaptation section, indicating that there is a certain deviation between the current process parameter combination and the batch characteristics, but it can be optimized through fine tuning. The system enters the process dynamic fine tuning state, starts the local parameter self-adaptive adjustment mechanism according to the real-time collected process characteristic data change trend, dynamically refines the key process parameter settings of reaction temperature, feed flow and catalyst ratio, improves the matching degree of batch characteristics and process parameters; At the same time, based on the characteristic vector change trend and the historical optimization path, a precise process optimization suggestion is generated to guide the operator to gradually adjust the control range, adjustment step and parameter coordination change strategy of the related parameters, dynamically correct the process settings to adapt to the batch characteristic change, real-time update the dynamic control response curve, ensure that the process parameters remain in the optimal response state in the multi-dimensional characteristic space, reduce the energy consumption fluctuation amplitude and improve the output quality stability;
[0067] When the dynamic optimization adjustment value is less than the second control threshold, it is determined that the low matching adaptation section, indicating that the current process parameter combination is seriously mismatched with the batch characteristics, and there is a process execution risk. The system will immediately suspend the process execution process of the current batch, start the emergency process rollback mechanism, and quickly switch to the predefined emergency process parameter template to ensure production continuity and system safety. At the same time, based on the current batch characteristic features, historical deviation patterns and abnormal distribution characteristics, the individualized optimization adjustment guidance for the abnormal batch is output, which clearly proposes the process parameter rollback path, key characteristic correction direction, process control response strategy and next step dynamic optimization suggestion, helping the operator to accurately correct the process parameter configuration in the shortest time, restore the production process stability, improve the real-time adaptability, process control robustness and abnormal recovery ability of the system to nonlinear complex characteristic batches, build a closed-loop dynamic optimization response mechanism, and support the intelligent, precise and efficient upgrading of the dimethyl sulfoxide production process.
[0068] In the embodiment, by comparing the dynamic optimization adjustment value with the process control response threshold in real time, the matching state of different process parameter configurations and batch characteristics in the production process is dynamically identified, and differential control strategies are executed according to high matching, medium matching and low matching sections. For the high matching section, the existing process parameter settings are maintained, the optimization guidance for maintaining process stability and trend monitoring is generated, and the process continuity and energy efficiency consistency are improved. For the medium matching section, a local parameter self-adaptive fine-tuning mechanism is started, the key process parameter settings are optimized based on the real-time characteristic change trend, and the dynamic matching degree and control precision of batch characteristics and process parameters are improved. For the low matching section, the current batch process is suspended, switched to the emergency process parameter template, and the abnormal batch correction guidance is output to quickly restore the process stability and production safety. Through this step, the dynamic response speed, abnormal response ability and system intelligence level of process optimization in the dimethyl sulfoxide production process are effectively improved, supporting the evolution of the production process to the direction of precision, self-adaptation and high efficiency.
[0069] The second aspect of the application provides a dimethyl sulfoxide production optimization system based on a deep neural network, comprising a data acquisition and preprocessing module, a parameter relationship analysis module, a sensitivity analysis module and a dynamic optimization module. The data acquisition and preprocessing module deploys multiple types of detection terminals to collect current batch process characteristic data, production environment data and processing process data, covering key characteristics such as raw material ratio, reaction temperature curve, reaction pressure curve, catalyst concentration, stirring rate change, raw material flow fluctuation, energy consumption index and product purity, while collecting environmental and process auxiliary characteristics such as workshop temperature, pressure, humidity, gas flow rate, equipment operating state and energy consumption load to ensure multi-dimensionality and full coverage of data. The collected process characteristic data, production environment data and processing process data are standardized and normalized for preprocessing, unified timestamp, cleaning of abnormal values and missing values, standardization of physical units and normalization of numerical value range, construction of structured and extensible standardized data set, and provision of high-quality data basis for subsequent characteristic modeling and optimization analysis. The parameter relationship analysis module, based on the standardized data set and combining historical batch process parameter characteristics, uses correlation analysis, nonlinear regression and adaptive fitting methods to comprehensively evaluate the nonlinear coupling relationship between different process parameters, identify the comprehensive influence of strongly coupled parameters on production process stability and energy efficiency changes, and generate process parameter interaction values for the current batch. At the same time, based on the interaction value, the subsequent data acquisition frequency is dynamically optimized to improve the response speed to sensitive parameter change trends, and the process monitoring priority and dynamic data acquisition strategy are adjusted according to the interaction strength to achieve simultaneous improvement of data acquisition accuracy and system response performance. The sensitivity analysis module performs process sensitivity analysis on each process parameter in the dimethyl sulfoxide production process under the current batch characteristics based on the standardized data set, quantifies the comprehensive influence of each process parameter on energy consumption changes, resource consumption and product yield fluctuations, comprehensively considers the historical fluctuation characteristics and real-time response trend of the process parameters, dynamically evaluates the sensitivity level of the parameters, and then updates the adjustment amplitude interval, parameter fine-tuning step and dynamic correction strategy of each key process parameter according to the process sensitivity analysis results to enhance the real-time and response accuracy of parameter control and ensure process stability and output consistency during optimization. The dynamic optimization module takes the current batch process parameter interaction value and process sensitivity analysis result as input, comprehensively analyzes the comprehensive dynamic matching degree of the current batch under nonlinear process characteristics and strongly coupled parameter characteristics, integrates the characteristic vector change trend and historical batch optimization experience, dynamically generates batch characteristic and process parameter optimal matching degree evaluation value, and deduces the optimal process adjustment optimization suggestion based on the comprehensive dynamic matching degree analysis result to clearly identify the process parameters that need to be fine-tuned first, the recommended adjustment range, the optimization control path and the dynamic response strategy, assisting operators in accurately adjusting process settings and improving the adaptive optimization capability and process control intelligent level of the production process.
[0070] In this embodiment, the data acquisition and preprocessing module is used to deploy multiple types of detection terminals to comprehensively collect process characteristic data, production environment data and processing process data of the current batch, covering multi-dimensional characteristics including raw material ratio, reaction parameters, equipment state and energy consumption indicators. Through standardized and normalized preprocessing, time synchronization, physical unit unification and numerical normalization are realized, abnormal values and missing values are cleaned up, and a structured, highly complete and consistent standardized data set is constructed. This module provides a high-quality data foundation for subsequent characteristic modeling and process optimization analysis, improving the accuracy, timeliness and scalability of data acquisition;
[0071] The parameter relationship analysis module is used to comprehensively evaluate the nonlinear coupling relationship between different process parameters based on the standardized data set combined with historical batch process characteristics, using correlation analysis and nonlinear fitting methods, to quantify the interactive influence of each parameter on process stability and energy efficiency changes, and to generate process parameter interaction values for the current batch. At the same time, according to the interaction strength, the data acquisition frequency and monitoring strategy are dynamically optimized to improve the response ability of the system to the process parameter change trend. This module supports the identification and modeling of complex relationships between process characteristics, providing a foundation for subsequent sensitivity analysis and dynamic optimization;
[0072] The sensitivity analysis module is used to perform process sensitivity analysis on the current batch process characteristics based on the standardized data set, to quantify the influence of each process parameter on energy consumption, resource consumption and output fluctuations, and to dynamically evaluate the sensitivity level of the parameters. Based on the sensitivity analysis results, the adjustment range and fine-tuning strategy of the key process parameters are updated in real time to optimize the parameter adjustment precision and response speed. This module improves the adaptive adjustment capability of process parameters, enhances the dynamic adaptability of the production process to batch characteristic changes and the fine-tuning level of process control;
[0073] The dynamic optimization module is used to input the process parameter interaction values and sensitivity analysis results, to comprehensively analyze the comprehensive dynamic matching degree of the current batch under nonlinear characteristics and strong coupling characteristics, to integrate historical optimization experience and characteristic vector change trend, and to derive optimal process adjustment optimization suggestions. It clearly defines the priority adjustment parameters, recommended adjustment range and dynamic control path to assist operators in accurately performing process optimization operations. This module realizes real-time adaptive optimization of process parameters in the production process, improving the intelligentization, precision and dynamic response ability of process control.
[0074] It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting; it is not intended to exclude myriad other embodiments of the present application that other presenters can develop. It is also possible, however, that only a single element can be present. It is further noted that such a term as "comprising" is intended to mean that the embodiments include the recited elements, but not excluding other elements. "Consisting essentially of when used herein in relation to a composition, means that the composition includes the recited elements, and can include additional elements, so long as the additional elements do not materially alter the basic and novel properties of the claimed composition. "Consisting of" when used herein in relation to a composition means that the composition includes the recited elements and nothing more.
[0075] The preferred embodiments of the application disclosed above are only to help explain the principles of the present application. The preferred embodiments do not limit the present application to only the specific embodiments described. It is apparent that many modifications and variations of this application are possible in light of this disclosure. The preferred embodiments are chosen and described in order to best explain the principles of the application and the practical application, to thereby enable others skilled in the art to best utilize the application. The present application is limited only by the claims and their full scope and equivalents.
Claims
1. A dimethyl sulfoxide production optimization method based on deep neural network, characterized in that: include: S1, deploy multiple types of detection terminals to collect the process characteristic data, production environment data and processing process data of the current batch, and standardize and normalize the collected process characteristic data, production environment data and processing process data to build a standardized data set; S2, based on the standardized data set and the characteristics of each process parameter in historical batches, comprehensively evaluates the nonlinear coupling relationship between different process parameters, generates the process parameter interaction value for the current batch, and optimizes the subsequent data collection frequency and the priority of the dynamic adjustment strategy based on the process parameter interaction value; S3, based on the standardized data set, conducts process sensitivity analysis on each process parameter in the dimethyl sulfoxide production process under the current batch characteristics, and then updates the adjustment range and dynamic correction strategy of each key process parameter according to the process sensitivity analysis results; S4, taking the interaction values of the process parameters of the current batch and the process sensitivity analysis results as input, comprehensively analyzes the comprehensive dynamic matching degree of the current batch under the nonlinear process characteristics and strong coupling parameter characteristics, and generates the optimal process adjustment optimization suggestions based on the comprehensive dynamic matching degree analysis results.
2. The method for optimizing dimethyl sulfoxide production based on a deep neural network according to claim 1, characterized in that: The specific steps of deploying multiple types of detection terminals to collect process characteristic data, production environment data, and processing data of the current batch are as follows: Collect process characteristic data including raw material ratio, reaction temperature curve, reaction pressure curve, catalyst concentration, stirring rate change, raw material flow fluctuation, reaction time, product yield, product purity and energy consumption indicators through production process detection equipment. At the same time, record the measurement unit, measurement method, detection accuracy, unit energy consumption change, detection equipment model, ratio of historical measurement mean to standard deviation, degree of deviation of current process characteristics from historical mean, instantaneous fluctuation amplitude and ambient temperature, humidity and air pressure conditions at the time of sampling for each process characteristic data; The process environment monitoring equipment is used to collect production environment data including the temperature, pressure, humidity, gas flow rate, heat exchange efficiency, equipment operation time, raw material consumption rate and condensation efficiency around the reactor. At the same time, the collection frequency, process operation frequency, change rate, real-time collection value, resource consumption, monitoring period, historical monthly average and standard deviation, current process section number and production batch number of each process parameter are recorded; The records include the total number of collected process parameters, timestamp of each collected data, sampling interval, measurement duration, collection data completeness rate, historical collection mean, unit energy consumption change mean, batch data coverage rate and environmental monitoring synchronization rate of processing data.
3. The method for optimizing dimethyl sulfoxide production based on a deep neural network according to claim 1, wherein: The collected process characteristic data, production environment data and processing process data are standardized and normalized preprocessed to construct a standardized data set. The specific steps are: Through a unified time synchronization mechanism, all process characteristic data and production environment data are time-stamped and standardized, and the time alignment success rate and time drift average of each production batch data are recorded; Clean the null values, invalid extreme values, format abnormal fields and measurement error exceeding the limit in the process characteristic data, production environment data and processing process data, and record the missing rate, abnormal proportion and rejection ratio of each process characteristic item; Unify the process characteristic field names output by different testing equipment into internal standard field specifications, retain the original field names and mapping relationships, and establish a standard field dictionary table; The process characteristic fields with complex structures are flattened and expanded according to unified rules, split into a one-dimensional process characteristic field set, and an index table associating the fields with the production batch numbers is established; Unify the physical unit representation of all collected fields, use the International System of Units for conversion, and record the unit conversion success rate, the proportion of unit abnormal fields, and the unit standardization completion rate; Normalize the collected process characteristic data, production environment data, and processing process data to eliminate the dimension and scale differences of process parameters between batches; The total number of sampling parameter types, the number of historical production batches, and the actual processing values of each process parameter for each batch of products, rule correction values, number of abnormal monitoring indicators, characteristic stability values and batch characteristic standardization completion rate are counted. The standardized and normalized process characteristic data, production environment data and processing process data are stored to build a standardized data set.
4. The method for optimizing dimethyl sulfoxide production based on a deep neural network according to claim 1, wherein: The specific steps of comprehensively evaluating the nonlinear coupling relationship between different process parameters based on the standardized data set and combining the process parameter characteristics of historical batches to generate the process parameter interaction value for the current batch are as follows: Extract the total number of collected process parameters and the real-time collected values and historical collected mean values of each process parameter of the current batch from the standardized database, calculate the difference between the real-time collected values and the historical collected mean values, and take the absolute value; extract the instantaneous fluctuation amplitude of each process parameter of the current batch at the current moment, divide the absolute value of the difference between the real-time collected values and the historical collected mean values by the instantaneous fluctuation amplitude to obtain the deviation value; perform product operation on the deviation values of all collected process parameters one by one to obtain the overall process deviation product; Extract the change rate of each process parameter at the current moment and add it to the stability factor to obtain the change stability value. Add the change stability values of all collected process parameters to obtain the overall change trend value. The overall process deviation product is divided by the overall change trend value and the logarithm is taken to obtain the process parameter interaction value of the current batch in the production process.
5. The method for optimizing dimethyl sulfoxide production based on a deep neural network according to claim 1, wherein: The specific steps for optimizing the subsequent data collection frequency and the priority of the dynamic adjustment strategy based on the process parameter interaction value are as follows: Real-time comparison of process parameter interaction values and interaction stability thresholds. The interaction stability thresholds include the first stability threshold and the second stability threshold: When the process parameter interaction value is greater than or equal to the first stability threshold, it is determined to be a high stability section. The current data acquisition and processing strategy remains unchanged, and the process optimization is directly derived based on the evaluation value without the need to introduce additional numerical compensation terms. When the process parameter interaction value is greater than or equal to the second stability threshold and less than the first stability threshold, it is determined to be in the medium stability section and enters the numerical warning state. A moderate compensation strategy is implemented, including dynamically adjusting the data acquisition frequency, expanding the sampling window range, and introducing small positive numbers into the change rate and fluctuation amplitude of each process parameter to enhance the numerical stability of the formula. At the same time, formula operation error monitoring is activated to ensure the continuity and reliability of the calculation process. When the interaction value of the process parameters is less than the second stability threshold, it is determined to be a low stability section. The process optimization derivation based on the current data is immediately suspended, the numerical anomaly emergency correction mechanism is activated, the historical batch process characteristics are re-extracted, and the dynamic weighted average parameters of multiple batches are constructed to replace the abnormal parameters of the current batch in the formula calculation. At the same time, the current batch is listed as a high-priority data quality traceability object, and the data deviation characteristics and stability failure conditions are recorded.
6. The method for optimizing dimethyl sulfoxide production based on a deep neural network according to claim 1, wherein: The specific steps of performing process sensitivity analysis on each process parameter in the dimethyl sulfoxide production process under the current batch characteristics based on the standardized data set are as follows: Extract the instantaneous fluctuation amplitude of each process parameter of the current batch under the current batch, the resulting unit energy consumption change, and the average unit energy consumption change of historical batches from the standardized database; calculate the absolute value of the difference between the unit energy consumption change of the current batch and the average unit energy consumption change of historical batches and divide it by the instantaneous fluctuation amplitude to obtain the normalized energy consumption value of each process parameter; Extract the process operation frequency and resource consumption of each process parameter in the current batch, divide the process operation frequency by the resource consumption coefficient, and obtain the resource utilization efficiency value of each process parameter; Multiply the normalized energy consumption value by the resource utilization efficiency value, add one, and then take the logarithm to obtain the process sensitivity evaluation value of the current process parameters.
7. The method for optimizing dimethyl sulfoxide production based on a deep neural network according to claim 1, wherein: The specific steps of updating the adjustment range and dynamic correction strategy of each key process parameter according to the process sensitivity analysis result are as follows: Compare the process sensitivity evaluation value with the sensitivity threshold in real time, where the sensitivity threshold includes a first sensitivity threshold and a second sensitivity threshold: When the process sensitivity evaluation value is greater than or equal to the first sensitivity threshold, it is determined to be a high-sensitivity section. The existing process parameter configuration and data collection frequency remain unchanged, the production process is continuously executed, and a high-priority data archiving mode is entered to record energy consumption indicators, raw material consumption curves and operation logs in real time. When the process sensitivity assessment value is greater than or equal to the second sensitivity threshold and less than the first sensitivity threshold, it is determined to be in the medium sensitivity section, triggering the parameter fine-tuning mechanism. Based on the real-time monitoring data trend, the reaction temperature, feed flow rate and catalyst addition ratio are dynamically adjusted. The sampling frequency is simultaneously increased, the energy consumption and raw material usage data collection cycle is shortened, and the monitoring density of energy efficiency sensitive parameters and key resource consumption indicators is enhanced. The intermediate energy efficiency risk warning function is activated to remind operators to pay attention to process fluctuations and prevent further expansion of energy efficiency anomalies. When the process sensitivity evaluation value is less than the second sensitivity threshold, it is determined to be a low-sensitivity section. The production operation based on the current parameters is immediately suspended, and the emergency parameter correction mechanism is activated. At the same time, the current batch is included in the high-priority energy efficiency anomaly tracing and risk assessment objects, and the energy consumption deviation characteristics, resource waste trends and abnormal operation logs are fully recorded.
8. The method for optimizing dimethyl sulfoxide production based on a deep neural network according to claim 1, wherein: The specific steps of comprehensively analyzing the comprehensive dynamic matching degree of the current batch under the nonlinear process characteristics and the strong coupling parameter characteristics using the process parameter interaction value and the process sensitivity analysis result of the current batch as input are as follows: Obtaining a process parameter interaction value and a process sensitivity evaluation value, raising the process sensitivity evaluation value to the power of a sensitivity adjustment factor and multiplying the result by the process parameter interaction value to obtain a process interaction sensitivity combination value; Extract the completeness rate of the collected data of the current process parameters, multiply the trend sensitivity coefficient by the absolute value of the second-order inverse of the collected data completeness rate, and sum it with the nonlinear adjustment exponent power of the absolute value of the first-order derivative of the collected data completeness rate to obtain the collected data sensitivity value; The process interaction sensitivity combination value is divided by the collected data sensitivity value and then added with one to take the logarithm to obtain the dynamic optimization adjustment value of the current process parameters.
9. The method for optimizing dimethyl sulfoxide production based on a deep neural network according to claim 1, wherein: The specific steps of generating the optimal process adjustment optimization suggestion based on the comprehensive dynamic matching degree analysis result are as follows: The dynamic optimization adjustment value is compared with the process control response threshold in real time, wherein the process control response threshold includes a first control threshold and a second control threshold: When the dynamic optimization adjustment value is greater than or equal to the first control threshold, it is determined to be a high-match adaptation section, and the current process parameter settings remain unchanged. At the same time, optimization suggestions for maintaining process stability are generated, and key parameters for subsequent monitoring and preventive adjustment guidelines are clarified to assist operators in continuously monitoring process fluctuation trends and ensure that the production process operates efficiently within a stable range. When the dynamic optimization adjustment value is greater than or equal to the second control threshold and less than the first control threshold, it is determined to be a medium matching adaptation section and enter the process dynamic fine-tuning state. According to the change trend of the process characteristic data collected in real time, the local parameter adaptive adjustment mechanism is started to dynamically refine the key process parameters of reaction temperature, feed flow rate and catalyst ratio, improve the matching degree between batch characteristics and process parameters, and generate process optimization suggestions based on the change trend of the characteristic vector to guide the operator to gradually adjust the control range and adjustment step of the relevant parameters, dynamically correct the process settings to adapt to the changes in batch characteristics, and ensure that the process parameters maintain the optimal response state in the multi-dimensional characteristic space; When the dynamic optimization adjustment value is less than the second control threshold, it is determined to be a low-matching adaptation section. The process execution flow of the current batch is immediately suspended, the emergency process fallback mechanism is activated, and the predefined emergency process parameter template is quickly switched to ensure production safety. At the same time, based on the characteristics of the current batch, optimization adjustment guidelines for abnormal batches are output, and the process parameter fallback path, key characteristic correction direction and next dynamic optimization strategy recommendations are clearly proposed to help operators restore process stability in the shortest time, improve the system's real-time adaptability and process control robustness to batches with nonlinear and complex characteristics, and build a closed-loop dynamic optimization response mechanism.
10. A dimethyl sulfoxide production optimization system based on deep neural network, characterized in that: include: Data acquisition and preprocessing module, parameter relationship analysis module, sensitivity analysis module and dynamic optimization module, including: The data collection and preprocessing module deploys multiple types of detection terminals to collect the process characteristic data, production environment data, and processing process data of the current batch, and standardizes and normalizes the collected process characteristic data, production environment data, and processing process data to build a standardized data set; The parameter relationship analysis module comprehensively evaluates the nonlinear coupling relationship between different process parameters based on standardized data sets and the characteristics of each process parameter in historical batches, generates process parameter interaction values for the current batch, and optimizes the subsequent data collection frequency and the priority of the dynamic adjustment strategy based on the process parameter interaction values; The sensitivity analysis module performs process sensitivity analysis on various process parameters in the dimethyl sulfoxide production process based on the standardized data set under the current batch characteristics, and then updates the adjustment range and dynamic correction strategy of each key process parameter based on the process sensitivity analysis results; The dynamic optimization module takes the interaction values of the current batch process parameters and the process sensitivity analysis results as input, comprehensively analyzes the comprehensive dynamic matching degree of the current batch under the nonlinear process characteristics and strong coupling parameter characteristics, and generates optimal process adjustment optimization suggestions based on the comprehensive dynamic matching degree analysis results.
Citation Information
Cited By
Semiconductor process improvement effect analysis method, system, equipment and medium
CN121285275A
MES (Manufacturing Execution System) for automatically adjusting server manufacturing process based on current network fault
CN121365309A
Mes system for automatically adjusting server manufacturing process based on network fault
CN121365309B
Multi-parameter cooperative control method and system for chemical product production workshop
CN121455109A
Natural gas hydrogen production reaction parameter optimization method and system
CN121687238A