Chemical change risk management and control method and system based on generative artificial intelligence

Abstract

The invention discloses a chemical engineering change risk management and control method and system based on generative artificial intelligence, and belongs to the technical field of risk management and control. Historical chemical engineering risk influence data in a chemical engineering production process are collected and preprocessed; the method comprises the steps of preprocessing historical chemical engineering risk influence data, training a chemical engineering change risk identification model by adopting the preprocessed historical chemical engineering risk influence data, collecting real-time chemical engineering risk influence data after the chemical engineering change when the chemical engineering change is generated, and identifying the real-time chemical engineering risk influence data by adopting the trained chemical engineering change risk identification model. And finally, the chemical change risk is managed and controlled according to the identified chemical change risk identification result, so that the monitoring efficiency of the chemical change risk can be effectively improved, the monitoring accuracy of the chemical change risk can also be improved, and the safety in the chemical production process is powerfully guaranteed.

Description

Technical Field

[0001] The present invention belongs to the technical field of risk control, and particularly relates to a chemical process change risk control method and system based on generative artificial intelligence. Background Art

[0002] As an important pillar of the national economy, the chemical industry involves various complex chemical reactions in its production process, presenting relatively high safety risks. Chemical process change risk refers to the potential safety hazards caused by changes in process, equipment, raw materials, or operating conditions during chemical production. These changes may affect key parameters such as reaction rate, temperature, and pressure, leading to uncontrollable chemical reactions or equipment failures, and further causing serious accidents such as leaks, fires, and explosions. The control of chemical process change risk is of crucial importance. It is necessary to identify, evaluate, and control potential risks through strict risk assessment, safety analysis, and approval procedures to ensure production safety. Effectively managing chemical process change risk is an important measure to ensure the safe production of the chemical industry, protect the environment, and safeguard the safety of personnel's lives and property. Traditional manual risk control methods rely on empirical judgment, suffering from problems such as strong subjectivity, low efficiency, and difficulty in comprehensive coverage. Therefore, there is an urgent need for an intelligent and efficient chemical process change risk control method. Summary of the Invention

[0003] The present invention provides a chemical process change risk control method and system based on generative artificial intelligence to solve the problems of strong subjectivity and low efficiency caused by manual risk control in the prior art.

[0004] On the one hand, the present invention provides a chemical process change risk control method based on generative artificial intelligence, including: Collecting historical chemical risk impact data during the chemical production process, and preprocessing the historical chemical risk impact data to obtain the preprocessed historical chemical risk impact data; Using a generative artificial intelligence model to construct a chemical process change risk identification model, and training the chemical process change risk identification model with the preprocessed historical chemical risk impact data to obtain the trained chemical process change risk identification model; When a chemical process change occurs, collecting the real-time chemical risk impact data after the chemical process change, and using the trained chemical process change risk identification model to identify the real-time chemical risk impact data to determine the chemical process change risk identification result; Based on the chemical process change risk identification result, controlling the chemical process change risk to complete the chemical process change risk control based on generative artificial intelligence.

[0005] Further, collecting historical chemical risk impact data during the chemical production process includes: Collect process parameters, equipment operation data, raw material properties, and environmental factors in the chemical production process to obtain historical chemical risk impact data in the chemical production process; among them, the historical chemical risk impact data includes historical chemical risk impact data under normal conditions and historical chemical risk impact data under risk conditions.

[0006] Further, preprocess the historical chemical risk impact data to obtain the preprocessed historical chemical risk impact data, including: Perform data cleaning, missing value filling, and outlier processing on the historical chemical risk impact data to obtain the preprocessed historical chemical risk impact data.

[0007] Further, use a generative artificial intelligence model to construct a chemical change risk identification model, including: using a generative adversarial network or a variational autoencoder to construct a chemical change risk identification model.

[0008] Further, use the preprocessed historical chemical risk impact data to train the chemical change risk identification model to obtain the trained chemical change risk identification model, including: Use the chaotic uniform mapping initialization method to initialize the network parameters of the chemical change risk identification model to obtain multiple different network parameter individuals; Use the preprocessed historical chemical risk impact data as input to obtain the loss function value corresponding to each parameter individual, and determine the network parameter individual with the smallest loss function value as the optimal parameter individual; Using the optimal parameter individual as a reference, use the optimal position siege strategy to perform local search on the network parameter individuals to obtain the network parameter individuals after local search; For the network parameter individuals after local search, use the neighborhood information exchange strategy to perform local area fusion search on the network parameter individuals to obtain the network parameter individuals after local area fusion search; For the network parameter individuals after local area fusion search, use the local jump search strategy to perform local area jump search on the network parameter individuals to obtain the network parameter individuals after local area jump search; Judge whether the current training times have reached the maximum training times. If so, re-determine the optimal parameter individual according to the network parameter individuals after local area jump search, and obtain the trained chemical change risk identification model with the re-determined optimal parameter individual. Otherwise, return to the step of obtaining the optimal parameter individual.

[0009] Further, use the chaotic uniform mapping initialization method to initialize the network parameters of the chemical change risk identification model to obtain multiple different network parameter individuals, including: Based on the upper and lower limits of the network parameters of the chemical process change risk identification model, randomly generate network parameters, encode the generated network parameters, and obtain network parameter individuals; Based on the obtained network parameter individuals, use chaotic mapping to obtain multiple different network parameter individuals as:

[0010] wherein, represents the nth network parameter individual, and when n = 1, represents the obtained network parameter individual, represents the chaotic mapping control parameter, represents the (n + 1)th network parameter individual.

[0011] Furthermore, taking the optimal parameter individual as a reference, use the optimal position siege strategy to perform local search on the network parameter individuals, and obtain the network parameter individuals after local search, including:

[0012]

[0013]

[0014] wherein, represents the t th network parameter individual in the i th training process, i = 1, 2,..., N, N represents the total number of network parameter individuals, represents the i th network parameter individual after local search, represents the random step size generated by Levy flight, represents a random number between (0, 1), represents the optimal parameter individual, represents a random number between (0, 1), represents another network parameter individual randomly matched for the network parameter individual, represents the first coefficient, which is randomly 0.01 or -0.01; represents the second coefficient, and as the iteration progresses, it linearly decreases from 1.9 to 0; represents the adaptive adjustment information term, represents the adaptive adjustment factor, represents a random number uniformly distributed within [0, 0.5], represents a random number between (0, 1), represents the preset maximum number of training times, represents the network parameter individual The Euclidean distance from the individual with optimal parameters is denoted by ||, which is the symbol for taking the absolute value of each element.

[0015] Furthermore, for the network parameter individuals after local search, a neighborhood information exchange strategy is adopted to perform local area fusion search on the network parameter individuals, obtaining the network parameter individuals after local area fusion search, including:

[0016] where, represents the t th network parameter individual after local search in the m th training process, represents the m th network parameter individual after local area fusion search, m = 1, 2,..., N, where N represents the total number of network parameter individuals, represents a random network parameter individual within the neighborhood range of the network parameter individual , and the neighborhood range is determined by the neighborhood radius R, represents a random angle between (0, 2 π ), u represents the first helix constant, v represents the second helix constant, e represents the natural constant, sin represents the sine function, and cos represents the cosine function.

[0017] Furthermore, for the network parameter individuals after local area fusion search, a local jump search strategy is adopted to perform local area jump search on the network parameter individuals, obtaining the network parameter individuals after local area jump search, including: For the network parameter individuals after local area fusion search, the local area jump search term corresponding to the network parameter individual is obtained as:

[0018] where, represents the t th network parameter individual after local area fusion search in the k th training process, represents the local area jump search term corresponding to the k th network parameter individual, k = 1, 2,..., N, where N represents the total number of network parameter individuals, represents the t th network parameter individual after local area fusion search in the q th training process, represents a randomly generated positive integer, and less than N, represents the individual with optimal parameters; Determine whether the loss function value of the local area jump search term decreases. If so, use the local area jump search term as the network parameter individual after the local area jump search; otherwise, directly use the network parameter individual after the original local area fusion search as the network parameter individual after the local area jump search.

[0019] On the other hand, the present invention provides a chemical process change risk control system based on generative artificial intelligence, including: a historical impact data collection module, a generative artificial intelligence training module, a chemical process change risk identification module, and a chemical process change risk control module; The historical impact data collection module is used to collect historical chemical risk impact data during the chemical production process, and preprocess the historical chemical risk impact data to obtain the preprocessed historical chemical risk impact data; The generative artificial intelligence training module is used to construct a chemical process change risk identification model using a generative artificial intelligence model, and train the chemical process change risk identification model using the preprocessed historical chemical risk impact data to obtain the trained chemical process change risk identification model; The chemical process change risk identification module is used to collect real-time chemical risk impact data after the chemical process change when a chemical process change occurs, and identify the real-time chemical risk impact data using the trained chemical process change risk identification model to determine the chemical process change risk identification result; The chemical process change risk control module is used to control the chemical process change risk based on the chemical process change risk identification result, and complete the chemical process change risk control based on generative artificial intelligence.

[0020] A chemical process change risk control method and system provided by the present invention collect historical chemical risk impact data during the chemical production process, preprocess the historical chemical risk impact data, train a chemical process change risk identification model using the preprocessed historical chemical risk impact data, collect real-time chemical risk impact data after the chemical process change when a chemical process change occurs, identify the real-time chemical risk impact data using the trained chemical process change risk identification model, and finally control the chemical process change risk based on the identified chemical process change risk identification result. This can not only effectively improve the monitoring efficiency of chemical process change risks, but also improve the monitoring accuracy of chemical process change risks, and effectively guarantee the safety during the chemical production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings here are incorporated into the specification and form a part of this specification, showing embodiments consistent with the present invention, and are used together with the specification to explain the principles of the present invention.

[0022] Figure 1 It is a flowchart of a chemical process change risk control method based on generative artificial intelligence provided for an embodiment of the present invention.

[0023] Figure 2 It is a schematic structural diagram of a chemical process change risk control system based on generative artificial intelligence provided for an embodiment of the present invention.

[0024] Among them, 201 - Historical impact data collection module, 202 - Generative artificial intelligence training module, 203 - Chemical process change risk identification module, 204 - Chemical process change risk control module.

[0025] Through the above accompanying drawings, specific embodiments of the present invention have been shown, and there will be more detailed descriptions hereinafter. These drawings and textual descriptions are not intended to limit the scope of the inventive concept in any way, but to illustrate the concept of the present invention to those skilled in the art by reference to specific embodiments. Detailed Description of the Invention

[0026] Here, the exemplary embodiments will be described in detail, and the examples are shown in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0027] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0028] First, the idea of the embodiments of the present invention will be introduced, which may include: training the generative artificial intelligence with a large amount of real chemical process change data to ensure that the model can accurately capture the key features in chemical process changes. After the training is completed, by inputting specific change parameters, such as process conditions, equipment types, raw material types, etc., the generative artificial intelligence can generate highly realistic simulated chemical process change scenarios. These simulated scenarios can include the process flow chart after the change, the equipment layout diagram, the reaction parameter change curve, etc., providing an intuitive and operable reference basis for chemical process change risk control. By analyzing these simulated scenarios, we can predict potential risk points and formulate corresponding safety measures, thereby effectively reducing the safety risks brought by chemical process changes.

[0029] As Figure 1 shown, the embodiments of the present invention provide a chemical process change risk control method based on generative artificial intelligence, including: S101. Collect historical chemical risk impact data during the chemical production process, and preprocess the historical chemical risk impact data to obtain the preprocessed historical chemical risk impact data; Historical chemical risk impact data are some data that can have an impact on risk types or risk events. After collecting the historical chemical risk impact data, it is also necessary to ensure that the data are all usable. Therefore, it is also necessary to preprocess the historical chemical risk impact data to ensure the accuracy of the collected data.

[0030] S102. Use a generative artificial intelligence model to construct a chemical change risk identification model, and use the preprocessed historical chemical risk impact data to train the chemical change risk identification model to obtain the trained chemical change risk identification model; Generative artificial intelligence models can efficiently create novel and personalized content, covering multiple fields such as text, images, and music. Their automated features reduce labor costs and improve production efficiency. The models optimize the output quality through continuous learning, demonstrating strong learning capabilities. At the same time, they provide inspiration for art, design, etc., and assist human creation. When data is scarce, they can generate simulated data to enhance the dataset. Their diversity, innovation, and broad application prospects are profoundly changing all walks of life. Therefore, the embodiments of the present invention use a generative artificial intelligence model to construct a chemical change risk identification model to achieve chemical change risk identification.

[0031] It should be noted that in the embodiments of the present invention, using a generative artificial intelligence model to construct a chemical change risk identification model is only a preferred implementation method. Other artificial intelligences can also be used to learn data relationships to achieve chemical change risk identification.

[0032] S103. When a chemical change occurs, collect the real-time chemical risk impact data after the chemical change, and use the trained chemical change risk identification model to identify the real-time chemical risk impact data to determine the chemical change risk identification result; After the chemical change risk identification model is trained, it has the ability to analyze data. Therefore, the trained chemical change risk identification model can be used to identify the real-time chemical risk impact data to determine whether the chemical change has risks, which can effectively assist users in improving the risk detection efficiency and accuracy.

[0033] S104. Based on the chemical change risk identification result, control the chemical change risk to complete the chemical change risk control based on generative artificial intelligence.

[0034] The results of chemical process change risk identification can be chemical scenarios such as the process flow diagram, equipment layout diagram, reaction parameter change curve, etc. after the change. Then, after obtaining the results of chemical process change risk identification, the results can be transmitted to the staff so that the staff can conduct risk control.

[0035] Furthermore, a deep learning model can be used to further identify the results of chemical process change risk identification obtained by simulation, determine the type or level of chemical risk, and then dispatch the preset risk control plan corresponding to the type or level of chemical risk, and forward the preset risk control plan to the staff for further processing to achieve chemical risk control. For example, a batch of training data can be prepared in advance, that is, the simulated results of chemical process change risk identification (i.e., simulated chemical process change scenarios), and then professional staff can label these results of chemical process change risk identification with classification labels, and a preset risk control plan is set for each classification label. Then, a deep learning model can be trained with the results of chemical process change risk identification and classification labels, and this trained deep learning model can achieve risk identification and thus find the corresponding risk control plan.

[0036] A chemical process change risk control method based on generative artificial intelligence provided by an embodiment of the present invention can effectively improve the efficiency and control ability of chemical risk control.

[0037] In an embodiment of the present invention, historical chemical risk impact data in the chemical production process is collected, including: Collect process parameters, equipment operation data, raw material properties, and environmental factors in the chemical production process to obtain historical chemical risk impact data in the chemical production process; among them, the historical chemical risk impact data includes historical chemical risk impact data under normal conditions and historical chemical risk impact data under risk conditions.

[0038] It should be noted that in order to enable generative artificial intelligence to generate corresponding scenarios, the chemical production scenarios corresponding to the historical chemical risk impact data should also be collected. These chemical production scenarios can include process flow diagrams, equipment layout diagrams, reaction parameter change curves, etc. The historical chemical risk impact data includes historical chemical risk impact data under normal conditions and historical chemical risk impact data under risk conditions. Therefore, the relationship between historical chemical risk impact data and chemical production scenarios can be continuously learned, so as to generate more realistic simulated chemical process change scenarios, discover potential risk points of chemical process changes, and improve the chemical process change risk control ability.

[0039] In an embodiment of the present invention, the historical chemical risk impact data is preprocessed to obtain the preprocessed historical chemical risk impact data, including: The historical chemical industry risk impact data is cleaned, missing values ​​are filled, and outliers are processed to obtain the historical chemical industry risk impact data after preprocessing.

[0040] Data cleaning may include: removing duplicate data; filling missing values ​​may include: filling with mean data in the same scenario, or directly removing data with missing values. Outlier processing may include: removing data with outliers.

[0041] In an embodiment of the present invention, a generative artificial intelligence model is used to construct a chemical change risk identification model, including: using a generative adversarial network or a variational autoencoder to construct a chemical change risk identification model.

[0042] Optionally, an embodiment of the present invention preferably adopts a generative adversarial network to construct a chemical change risk identification model, and uses the powerful data generation capability of the generative adversarial network to simulate the corresponding scenarios under the chemical risk impact data, thereby exploring potential risk points.

[0043] In an embodiment of the present invention, the pre-processed historical chemical risk impact data is used to train the chemical change risk identification model, and the trained chemical change risk identification model is obtained, including: The chaotic uniform mapping initialization method is used to initialize the network parameters of the chemical change risk identification model to obtain multiple different network parameter individuals; Taking the pre-processed historical chemical risk impact data as input, the loss function value corresponding to each parameter individual is obtained, and the network parameter individual with the smallest loss function value is determined as the optimal parameter individual; For example, for any individual network parameter, after applying the hyperparameters contained in the individual network parameter to the chemical change risk identification model, the preprocessed historical chemical risk impact data is used as the input of the generator to obtain the output of the generator (i.e., simulated chemical change scenario). Then, based on the simulated chemical change scenario and the chemical production scenario corresponding to the historical chemical risk impact data, the discriminator is used for identification and the loss function value is obtained.

[0044] Taking the optimal parameter individual as a reference, the optimal position siege strategy is used to perform local search on the network parameter individual to obtain the network parameter individual after local search; For the network parameter individuals after local search, a neighborhood information exchange strategy is used to perform local area fusion search on the network parameter individuals to obtain the network parameter individuals after local area fusion search; For the network parameter individuals after the local area fusion search, a local jump search strategy is used to perform a local area jump search on the network parameter individuals to obtain the network parameter individuals after the local area jump search; Determine whether the current number of training times has reached the maximum number of training times. If so, re-determine the optimal parameter individual according to the network parameter individuals after local area jump search, and obtain the chemical process change risk identification model after training with the re-determined optimal parameter individual. Otherwise, return to the step of obtaining the optimal parameter individual.

[0045] In the prior art, the training method of the generative adversarial network mainly relies on the method of gradient backpropagation for training, which will cause the optimization of hyperparameters to fall into local optima prematurely, resulting in the generative adversarial network being unable to effectively simulate real scenarios, and ultimately leading to the inability to accurately analyze chemical process change risks. Therefore, the embodiments of the present invention provide a training algorithm to improve the shortcomings of the prior art and enhance the accuracy of identifying chemical process change risks.

[0046] In the embodiments of the present invention, the chaotic uniform mapping initialization method is used to initialize the network parameters of the chemical process change risk identification model, and multiple different network parameter individuals are obtained, including: Based on the upper and lower limits of the network parameters of the chemical process change risk identification model, randomly generate network parameters, and encode the generated network parameters to obtain network parameter individuals; Based on the already obtained network parameter individuals, use chaotic mapping to obtain multiple different network parameter individuals as:

[0047] Among them, represents the nth network parameter individual, and when n = 1, represents the already obtained network parameter individual, represents the chaotic mapping control parameter, represents the (n + 1)th network parameter individual.

[0048] Through the chaotic uniform mapping initialization method in the embodiments of the present invention, the initial solutions can be more evenly distributed in the solution space, which can effectively improve the training speed of the algorithm and the ability to find the global optimal solution.

[0049] In the embodiments of the present invention, taking the optimal parameter individual as a reference, the optimal position siege strategy is used to perform local search on the network parameter individuals to obtain the network parameter individuals after local search, including:

[0050]

[0051]

[0052] Among them, represents the t th training process and thei An individual network parameter, i where \(i = 1,2,\cdots,N\), and \(N\) represents the total number of individual network parameters, represents the i individual network parameter after the \(i\)-th local search, represents the random step size generated by Levy flight, represents a random number between \((0,1)\), represents the optimal parameter individual, represents a random number between \((0,1)\), represents another individual network parameter randomly matched for the individual network parameter, represents the first coefficient, which is randomly \(0.01\) or \(-0.01\); represents the second coefficient, and it linearly decreases from \(1.9\) to \(0\) as the iteration progresses; represents the adaptive adjustment information item, represents the adaptive adjustment factor, represents a random number uniformly distributed within \([0,0.5]\), represents a random number between \((0,1)\), represents the preset maximum number of training times, represents the individual network parameter and the optimal parameter individual the Euclidean distance between them, and \(|\cdot|\) represents the symbol of taking the absolute value of each element.

[0053] The optimal position siege strategy provided by the embodiments of the present invention can be based on the optimal position, combined with other parameter individuals, to achieve adaptive search area selection, and at the same time incorporate Levy flight, which can maintain the global search ability of the algorithm in the early and middle stages. As the algorithm progresses, it can also gradually increase the convergence accuracy and comprehensively improve the optimization ability of the algorithm.

[0054] In the embodiments of the present invention, for the individual network parameter after local search, a neighborhood information exchange strategy is adopted to perform local area fusion search on the individual network parameter to obtain the individual network parameter after local area fusion search, including:

[0055] Among them, represents the t individual network parameter after the \(i\)-th local search in the \(t\)-th training process, m where \(i = 1,2,\cdots,N\), and \(N\) represents the total number of individual network parameters, represents the m individual network parameter after the \(j\)-th local area fusion search, m where \(j = 1,2,\cdots,N\), and \(N\) represents the total number of individual network parameters, represents in the individual network parameter Random network parameter individuals within the neighborhood range, where the neighborhood range is determined by the neighborhood radius R. represents a random angle between (0, 2 π ) u represents the first helix constant. v represents the second helix constant. e represents the natural constant, sin represents the sine function, and cos represents the cosine function.

[0056] The neighborhood information exchange strategy provided by the embodiments of the present invention can effectively develop the neighborhood in a spiral search manner, effectively explore unfamiliar areas, and improve the ability to search for the global optimal solution.

[0057] In the embodiments of the present invention, for the network parameter individuals after local area fusion search, a local jump search strategy is adopted to perform local area jump search on the network parameter individuals, and the network parameter individuals after local area jump search are obtained, including: For the network parameter individuals after local area fusion search, obtain the local area jump search term corresponding to the network parameter individual as:

[0058] where represents the t th network parameter individual after local area fusion search in the k th training process. represents the local area jump search term corresponding to the k th network parameter individual. k = 1, 2,..., N, where N represents the total number of network parameter individuals. represents the t th network parameter individual after local area fusion search in the q th training process. represents a randomly generated positive integer, and is less than N. represents the optimal parameter individual. Judge whether the loss function value of the local area jump search term decreases. If so, use the local area jump search term as the network parameter individual after local area jump search; otherwise, directly use the original network parameter individual after local area fusion search as the network parameter individual after local area jump search.

[0059] The local jump search strategy provided by the embodiments of the present invention can provide a strong ability to jump out of the local optimum, assist the algorithm in global optimization, and improve the training effect of the algorithm. And a greedy strategy is introduced to ensure the rapid convergence of the algorithm.

[0060] The present invention enhances the data learning effect by providing a new training algorithm, and finally improves the ability to identify chemical process change risks.

[0061] A chemical process change risk control method based on generative artificial intelligence provided by the present invention collects historical chemical risk impact data in the chemical production process, preprocesses the historical chemical risk impact data, and uses the preprocessed historical chemical risk impact data to train a chemical process change risk identification model. When a chemical process change occurs, real-time chemical risk impact data after the chemical process change is collected, and the trained chemical process change risk identification model is used to identify the real-time chemical risk impact data. Finally, the chemical process change risk is controlled based on the identified chemical process change risk identification result, which can not only effectively improve the monitoring efficiency of chemical process change risks, but also improve the monitoring accuracy of chemical process change risks, and effectively guarantee the safety in the chemical production process.

[0062] As Figure 2 shown, an embodiment of the present invention provides a chemical process change risk control system based on generative artificial intelligence, including: a historical impact data collection module 201, a generative artificial intelligence training module 202, a chemical process change risk identification module 203, and a chemical process change risk control module 204; The historical impact data collection module 201 is used to collect historical chemical risk impact data in the chemical production process, preprocess the historical chemical risk impact data, and obtain the preprocessed historical chemical risk impact data; The generative artificial intelligence training module 202 is used to construct a chemical process change risk identification model using a generative artificial intelligence model, and use the preprocessed historical chemical risk impact data to train the chemical process change risk identification model to obtain a trained chemical process change risk identification model; The chemical process change risk identification module 203 is used to collect real-time chemical risk impact data after the chemical process change when a chemical process change occurs, and use the trained chemical process change risk identification model to identify the real-time chemical risk impact data to determine the chemical process change risk identification result; The chemical process change risk control module 204 is used to control the chemical process change risk based on the chemical process change risk identification result to complete the chemical process change risk control based on generative artificial intelligence.

[0063] A chemical process change risk control system based on generative artificial intelligence provided by an embodiment of the present invention can execute the above method technical solution, and its principle and beneficial effects are similar, so details are not described here again.

[0064] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) that contain computer-usable program code.

[0065] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0066] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing devices to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0067] These computer program instructions can also be loaded onto a computer or other programmable data processing devices, such that a series of operation steps are performed on the computer or other programmable devices to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable devices provide steps for implementing the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0068] Those of ordinary skill in the art can understand that all or part of the steps of implementing the above-mentioned methods and facts can be completed by a program instructing relevant hardware. The program involved or the said program can be stored in a computer-readable storage medium. When the program is executed, it includes the following steps: At this time, the corresponding method steps are introduced. The storage medium can be ROM / RAM, magnetic disk, optical disk, etc.

[0069] The specific embodiments described above further elaborate on the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only the specific embodiments of the present invention and is not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A chemical change risk management method based on generative artificial intelligence, characterized in that: include: Collect historical chemical risk impact data in the chemical production process, and pre-process the historical chemical risk impact data to obtain the pre-processed historical chemical risk impact data; A generative artificial intelligence model is used to build a chemical change risk identification model, and the pre-processed historical chemical risk impact data is used to train the chemical change risk identification model to obtain the trained chemical change risk identification model; When a chemical change occurs, real-time chemical risk impact data after the chemical change is collected, and the trained chemical change risk identification model is used to identify the real-time chemical risk impact data to determine the chemical change risk identification result; Based on the chemical change risk identification results, the chemical change risks are managed and controlled, and chemical change risk management based on generative artificial intelligence is completed.

2. The chemical change risk management method based on generative artificial intelligence according to claim 1 is characterized in that: Collect historical chemical risk impact data during chemical production, including: Collect process parameters, equipment operation data, raw material properties and environmental factors in the chemical production process to obtain historical chemical risk impact data in the chemical production process; among them, historical chemical risk impact data includes historical chemical risk impact data under normal conditions and historical chemical risk impact data under risk conditions.

3. The chemical change risk management method based on generative artificial intelligence according to claim 1 is characterized in that: Preprocessing the historical chemical industry risk impact data to obtain the preprocessed historical chemical industry risk impact data, including: The historical chemical industry risk impact data is cleaned, missing values ​​are filled, and outliers are processed to obtain the historical chemical industry risk impact data after preprocessing.

4. The chemical change risk management method based on generative artificial intelligence according to claim 1 is characterized in that: A generative artificial intelligence model is used to build a chemical change risk identification model, including: using a generative adversarial network or a variational autoencoder to build a chemical change risk identification model.

5. The method for chemical change risk management based on generative artificial intelligence according to claim 4 is characterized in that: The pre-processed historical chemical risk impact data is used to train the chemical change risk identification model to obtain the trained chemical change risk identification model, including: The chaotic uniform mapping initialization method is used to initialize the network parameters of the chemical change risk identification model to obtain multiple different network parameter individuals; Taking the pre-processed historical chemical risk impact data as input, the loss function value corresponding to each parameter individual is obtained, and the network parameter individual with the smallest loss function value is determined as the optimal parameter individual; Taking the optimal parameter individual as a reference, the optimal position siege strategy is used to perform local search on the network parameter individual to obtain the network parameter individual after local search; For the network parameter individuals after local search, a neighborhood information exchange strategy is used to perform local area fusion search on the network parameter individuals to obtain the network parameter individuals after local area fusion search; For the network parameter individuals after the local area fusion search, a local jump search strategy is used to perform a local area jump search on the network parameter individuals to obtain the network parameter individuals after the local area jump search; Determine whether the current number of training times has reached the maximum number of training times. If so, redetermine the optimal parameter individual based on the network parameter individual after the local area jump search, and use the redetermined optimal parameter individual to obtain the chemical change risk identification model after training, otherwise return to the step of obtaining the optimal parameter individual.

6. The method for chemical change risk management based on generative artificial intelligence according to claim 5 is characterized in that: The chaotic uniform mapping initialization method is used to initialize the network parameters of the chemical change risk identification model to obtain multiple different network parameter individuals, including: Based on the upper and lower limits of the network parameters of the chemical change risk identification model, the network parameters are randomly generated, and the generated network parameters are encoded to obtain the network parameter individuals; Based on the network parameter individuals that have been obtained, chaotic mapping is used to obtain multiple different network parameter individuals: in, represents the nth network parameter individual, and when n=1, Indicates the individual network parameters that have been obtained. represents the chaotic mapping control parameter, Represents the n+1th network parameter individual.

7. The method for chemical change risk management based on generative artificial intelligence according to claim 6 is characterized in that: Taking the optimal parameter individual as a reference, the optimal position siege strategy is used to perform local search on the network parameter individual, and the network parameter individual after local search is obtained, including: in, Indicates t During the training i individual network parameters, i =1,2,...,N, where N represents the total number of network parameter individuals. Indicates i After local search, the network parameter individuals are represents the random step length generated by Levy flight, Represents a random number between (0,1), represents the optimal parameter individual, Represents a random number between (0,1), Represents other network parameter individuals that are randomly matched by network parameter individuals, represents the first coefficient, which is randomly 0.01 or -0.01; represents the second coefficient, and decreases linearly from 1.9 to 0 as the iteration proceeds; Represents an adaptive adjustment information item, represents the adaptive adjustment factor, represents a random number uniformly distributed within [0,0.5], Represents a random number between (0,1), Indicates the preset maximum number of training times. Represents individual network parameters With the optimal parameter individual The Euclidean distance between them, || represents the sign of taking the absolute value of each element.

8. The method for chemical change risk management based on generative artificial intelligence according to claim 7 is characterized in that: For the network parameter individuals after local search, the neighborhood information exchange strategy is used to perform local area fusion search on the network parameter individuals, and the network parameter individuals after local area fusion search are obtained, including: in, Indicates t During the training m After local search, the network parameter individuals are Indicates m The network parameter individuals after the local area fusion search, m =1,2,...,N, where N represents the total number of network parameter individuals. Indicates that the network parameters are The random network parameter individuals in the neighborhood range of , the neighborhood range is determined by the neighborhood radius R, means (0,2 π ), u represents the first helical constant, v represents the second helical constant, e represents a natural constant, sin represents the sine function, and cos represents the cosine function.

9. The method for chemical change risk management based on generative artificial intelligence according to claim 8 is characterized in that: For the network parameter individuals after the local area fusion search, a local jump search strategy is used to perform a local area jump search on the network parameter individuals, and the network parameter individuals after the local area jump search are obtained, including: For the network parameter individuals after the local area fusion search, the local area jump search items corresponding to the network parameter individuals are obtained as follows: in, Indicates t During the training k The network parameter individuals after the local area fusion search, Indicates k The local area jump search items corresponding to the individual network parameters, k =1,2,...,N, where N represents the total number of network parameter individuals. Indicates t During the training q The network parameter individuals after the local area fusion search, represents a randomly generated positive integer, and Less than N, represents the optimal parameter individual; Determine whether the loss function value of the local area jump search item is reduced. If so, the local area jump search item is used as the network parameter individual after the local area jump search. Otherwise, the network parameter individual after the original local area fusion search is directly used as the network parameter individual after the local area jump search.

10. A chemical change risk management and control system based on generative artificial intelligence, characterized in that: include: Historical impact data collection module, generation of artificial intelligence training module, chemical change risk identification module and chemical change risk management and control module; The historical impact data collection module is used to collect historical chemical risk impact data in the chemical production process, and pre-process the historical chemical risk impact data to obtain the historical chemical risk impact data after pre-processing; The generative artificial intelligence training module is used to construct a chemical change risk identification model using a generative artificial intelligence model, and train the chemical change risk identification model using pre-processed historical chemical risk impact data to obtain the trained chemical change risk identification model; The chemical change risk identification module is used to collect real-time chemical risk impact data after the chemical change when a chemical change occurs, and use the trained chemical change risk identification model to identify the real-time chemical risk impact data to determine the chemical change risk identification result; The chemical change risk management and control module is used to manage the chemical change risks based on the chemical change risk identification results, and complete the chemical change risk management and control based on generative artificial intelligence.

Images