New method for identifying component coupling relationship in chemical reaction model

Abstract

The invention belongs to the field of numerical simulation of combustion discipline, and relates to a novel method for identifying a component coupling relation in a chemical reaction model. The method is based on a data mining theory, takes a reliable simplified reaction model library as a basis, can accurately recognize the coupling relationship between different components, and remarkably reduces the scale of a final simplified reaction model. Meanwhile, based on the method, the influence of pressure and equivalence ratio conditions and simplified targets (the ignition delay period and the component concentration) on the simplified reaction model structure is explored, simplified working conditions and simplified targets are defined, and the calculation cost of large-scale reaction model simplification can be remarkably reduced.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of numerical simulation of combustion science, and relates to a new method for identifying coupling relationships of components in a chemical reaction model. BACKGROUND

[0002] A detailed chemical reaction model can accurately reproduce the ignition and oxidation processes of a fuel within a wide operating condition range. However, with the deepening of chemical kinetics research, the size of the reaction model is increasingly large, and the demand for computing resources also increases dramatically. By simplifying the detailed reaction model, the computing cost can be significantly reduced without sacrificing the specific prediction performance of the detailed reaction model, thereby realizing efficient and reliable combustion simulation. Although various methods for simplifying detailed reaction models have been proposed, the existing methods usually simplify the reaction model based on the influence of a single component or reaction on the target prediction value, ignoring the coupling relationships of different components / reactions in a complex chemical reaction model, which leads to a mutual restriction between the size and efficiency of the final simplified reaction model, and makes it difficult to accurately predict the combustion and emission characteristics of an actual fuel within a wide operating condition range. To solve this problem, there is an urgent need for a detailed chemical reaction model simplification method that can accurately capture the coupling relationships of different components or reactions, so that the final simplified reaction model can achieve high accuracy and high efficiency at the same time. SUMMARY

[0003] To construct a simplified reaction model with high accuracy and small size, the application proposes a new method for identifying coupling relationships of components in a chemical reaction model.

[0004] The technical scheme adopted by the application to solve the technical problem is that the application is a new method for identifying coupling relationships of components in a chemical reaction model, comprising the following steps: Step 1: Construction of a reliable simplified reaction model library A number of components are randomly deleted from a detailed chemical reaction model by an exhaustive method, and the simplified reaction models are screened based on the relative error of the simplified reaction models and the detailed reaction model with respect to the target prediction value, so as to obtain a reliable simplified reaction model library, which is converted into a binary matrix to identify the components of each model; Step 2: Analysis of coupling relationships and importance of components The relative error of the simplified model is taken as an eigenvalue of the matrix constructed in step 1, and the existence probability of component A in the reaction model can be obtained by combining the Gaussian distribution and the conditional probability formula; based on the state function curve, the error corresponding to the existence probability of 99.99% of the component can be used to determine the importance of different components; at the same time, a virtual component I is introduced, and the strength of the coupling relationship of the components contained in the virtual component I is determined by evaluating the importance of the virtual component I ; Step 3: Simplification target determination Using the method of step 2, the detailed reaction model is simplified for different pressures, different equivalence ratios and different simplification targets (ignition delay period and component concentration) under wide temperature conditions, respectively; by comparing the structures of the simplified reaction models under different conditions, the effects of pressure, equivalence ratio and simplification target on the structure of the simplified reaction model are summarized, and the simplification conditions and simplification targets are determined; Step 4: Construction of the final simplified reaction model Using the simplification conditions and simplification targets determined in step 3, the importance of components in the detailed reaction model is sorted using the method of step 2; then, based on the sorting, components are deleted one by one until the prediction value of the simplified reaction model under any condition exceeds the prediction uncertainty space of the detailed reaction model, obtaining an initial simplified model; finally, using a multi-objective genetic algorithm, the reaction rate constants in the fuel sub-model are optimized within the uncertainty range to obtain the final simplified reaction model.

[0005] Further, 2 to 5 components are randomly deleted from the detailed reaction model each time by the exhaustive method.

[0006] Further, the components remaining in the simplified reaction model are set to "1", and the deleted components are set to "0", the total number of components in the detailed chemical reaction model is m , and the total number of simplified reaction models in the simplified reaction model library is n， Then the simplified reaction model library is converted into a m × n matrix.

[0007] Further, combined with the Gaussian distribution and the conditional probability formula, according to the support degree S( A ) of component A , the expected value σ of the normal distribution, the relative error normal distribution variance A when component μ is removed, and the relative error normal distribution variance A when component μ exists, the existence probability function of component A in the reaction model can be obtained, and the formula is: Further, the virtual component I is used to determine the combination of the research components, if all components in I exist, the state of I is "1", otherwise it is "0"; through the state I of the virtual component Y I , the jThe state of each component Y j The number of components in the combination J， Determine virtual components I The formula is: By evaluating virtual components I The importance of virtual components can be understood to obtain the coupling relationship between different components; virtual components I The more important, the more virtual components I The stronger the coupling relationship between the components contained therein.

[0008] Furthermore, the detailed chemical reaction model is as follows: α -Methylnaphthalene reaction model.

[0009] Compared with the prior art, the beneficial effects of the present invention are: (1) The present invention can accurately capture the coupling relationship between different components and apply it to the simplification of reaction model. Compared with the traditional reaction model simplification technology, the final simplified model obtained by using the present invention is more compact and can significantly reduce the calculation time of multidimensional combustion simulation.

[0010] (2) The present invention can analyze the influence of pressure, equivalence ratio and simplification target on the structure of simplified reaction model, clarify the simplified operating conditions and simplification target, and the simplified reaction model obtained with narrow operating conditions and few targets can well predict the ignition, oxidation and combustion characteristics of fuel under wide operating conditions and multiple targets. It can significantly shorten the time of simplification of detailed reaction model and can be used to simplify large-scale detailed chemical reaction model. Attached Figure Description

[0011] Figure 1 A flowchart is provided to simplify the large-scale detailed chemical reaction model of α-methylnaphthalene using this method.

[0012] Figure 2 This is a detailed step diagram of the state probability function method.

[0013] Figure 3 A comparison of the predicted and experimental results of the simplified reaction model for α-methylnaphthalene during the burn-out period in a shock tube. Detailed Implementation

[0014] Unless otherwise specified, the technical or scientific terms used in this specification shall have the ordinary meaning that would be understood by one of ordinary skill in the art. The use of terms such as "first," "second," etc., in this specification does not indicate a specific order, quantity, or importance, but is used only to distinguish different objects or steps and to avoid confusion. Unless the context otherwise requires, "several" means "at least two"; "including" shall be understood as an open-ended expression, meaning "including but not limited to."

[0015] The methods, steps, modules, and functions described in this specification are only for illustrating the technical solutions of the present invention and are not intended to limit the scope of protection. Those skilled in the art, after reading this specification, may make equivalent substitutions or modifications to the method flow, model structure, or implementation without departing from the core ideas of the present invention, and such substitutions or modifications should be considered to fall within the scope of protection of this specification.

[0016] Example 1 The following is in conjunction with the appendix Figure 1 - Figure 3 The present invention will be described in detail.

[0017] A novel method for identifying component coupling relationships in chemical reaction models, which is used to analyze and simplify... α A detailed chemical reaction model for methylnaphthalene, with specific steps as follows: Step 1: α Construction of a reliable simplified reaction model library for methylnaphthalene from α In the detailed chemical reaction model of methylnaphthalene, 2–5 components are randomly removed to obtain a simplified reaction model, and the relative error (RE) of the simplified target (QoIs) is calculated by formula (3). In the formula, and The simplified model and the initial model are respectively in the 1st... Under the first working condition, for the first indivual The prediction results; for The number of [items / components] is determined. Based on this error, simplified reaction models are filtered using a preset threshold. The threshold is the prediction uncertainty space of the detailed reaction model, obtained using the SOBOL method. If the predicted value of the simplified reaction model exceeds the prediction uncertainty space of the detailed reaction model, the simplified reaction model is removed. This process is repeated to obtain a reliable library of simplified reaction models. To identify components in the simplified reaction models, retained components are set to "1", and deleted components are set to "0". The simplified reaction model library is then transformed into a [system / system]. m × n A Boolean matrix of dimension, where the total number of components in a detailed chemical reaction model is . m The total number of simplified reaction models in the simplified reaction model library is n .

[0018] Step 2: Analysis of component coupling relationships and component importance The relative error of each simplified reaction model in the simplified reaction model library is obtained through equation (3). Using the relative error as the eigenvalue, the frequency of different relative errors in the model library is statistically analyzed to obtain the distribution of the eigenvalues. With an interval of 0.001 relative error, the distribution of the eigenvalues ​​follows a normal distribution. Combined with the conditional probability formula, we can obtain: remember Formula (4) can be transformed into: Known conditions and Substituting into equation (7), we get: For a given group A If it contains n A simplified model library k If a simplified model is deleted (i.e., the state is 0), then the components... A The relevant parameter expressions can be calculated using the following formula: in, It is the first in the model library i The relative error of a simplified model; It is the mean of all relative errors. Accordingly, the components... A support and k It has the following relationship: According to formula (12), the trend of the existence probability of any component in a simplified reaction model as a function of relative error can be predicted from the known model library, thus providing an assessment method for measuring the importance of a component. It is assumed that when the state probability value of a component is close to 1, the component must exist. In this case, the smaller the absolute value of the relative error corresponding to the state probability function, the more important the component is.

[0019] To evaluate the coupling relationship among multiple components, a virtual component is introduced. I This virtual component represents a combination of multiple components. In a simplified reaction model, when these components are present simultaneously, I The state is "1" otherwise "0", i.e., virtual component. I The state is calculated using equation (13). in,Y I Representing virtual components I state, Y j Indicates the first in the combination j The state of each component J This indicates the number of components in the combination. The evaluation of virtual components... I By understanding the importance of this, the coupling relationship between multiple components can be obtained.

[0020] The importance of the components is ranked according to the above method. Based on this ranking, reactions are gradually removed, and the maximum relative error of the simplified reaction model in predicting the simplified target is calculated. The initial simplified reaction model is obtained by using the prediction uncertainty space of the detailed reaction model as a threshold.

[0021] Step 3: Simplify Target Definition Using the method in step 2, a simplified detailed flowchart of the detailed chemical reaction model is shown below. Figure 2 As shown, the main steps are summarized below: ① Calculate the standard deviation of the relative error of the simplified reaction model library using formula (11). .

[0022] ② Scan the simplified reaction model library to confirm the first k The row numbers of the components that are retained.

[0023] ③ Based on the result of step ②, calculate the first... k Support of each component .

[0024] ④ Based on the results of steps ② and ③, use formulas (9) and (10) to calculate the first... k Components and .

[0025] ⑤ Calculate using formulas (4) and (5) and And using formula (11) to obtain the first k The state probability function of each component.

[0026] ⑥ Calculate the relative error corresponding to the preset probability, and determine the importance of the component based on the relative error.

[0027] ⑦ Repeat steps ②–⑥ until the importance of all components has been assessed.

[0028] ⑧ Based on the results of step ⑦, rank the importance of the components.

[0029] ⑨ Based on the component sorting in step ⑧, delete components one by one until the predicted value of the simplified reaction model for any simplified target exceeds the prediction uncertainty space of the detailed reaction model, thus obtaining the final simplified mechanism.

[0030] To determine the simplified operating conditions and simplification objectives, the detailed chemical reaction model was simplified using the above method for different pressures (1–20 atm), different equivalence ratios (0.2–4.0), and different simplification objectives (ignition delay and component concentration) under wide temperature conditions. The specific operating conditions and objectives are shown in Table 1.

[0031] Table 1. Simplification objectives and operating conditions of the reaction model simplification Table 2 shows a comparison of the components in the final simplified reaction model under different pressures, with the flame delay period as the target. It can be seen that... p All components of the simplified reaction model obtained under the condition of 20 atm were included. p In the simplified reaction model obtained under the condition of =1 atm, i.e. based on p The simplified reaction model obtained under the condition of =1 atm can reproduce the reaction well. p Predictive performance of the detailed reaction model under a 20 atm condition.

[0032] Table 2 Comparison of component results in simplified reaction models targeting ignition delay under different pressures. Table 3 shows a comparison of the components in the final simplified reaction model under different equivalence ratios with the flame delay period as the target. It can be seen that... φ The simplified reaction model obtained under the 4.0 equivalence ratio condition includes all components of the simplified reaction models obtained under other equivalence ratio conditions.

[0033] Table 3 Comparison of component results in simplified reaction models targeting ignition delay under different equivalence ratios. Table 4 shows a comparison of the components in the simplified reaction models targeting ignition delay and component concentration, respectively. It can be seen that the simplified reaction model obtained with component concentration as the target can simultaneously and well reproduce the prediction characteristics of the detailed reaction model for ignition delay and component concentration.

[0034] Table 4 Comparison of Component Results of Simplified Reaction Models with Different Simplification Objectives In summary, the simplification of the reaction model should be based on a wide temperature range, p =1 atm and φ The component concentration under the condition of 4.0 is simplified to the target.

[0035] Step 4: Final Simplified Reaction Model Construction Based on the results of step 3, a simplification objective is selected, and the method of step 2 is used to refine the details. α The chemical reaction model for methylnaphthalene was simplified, reducing the detailed model containing 313 components and 2148 reactions to a simplified model with only 53 components and 262 reactions. Subsequently, a multi-objective genetic algorithm was used to optimize the reaction rate constants in the fuel sub-model, obtaining the final simplified reaction model, as shown below. Figure 3 As shown, the results indicate that the current simplified reaction model can reproduce the reaction well. α Ignition properties of methylnaphthalene.

Claims

1. A novel method for identifying component coupling relationships in a chemical reaction model, characterized in that, Includes the following steps: Step 1: Construction of a Reliable and Simplified Reaction Model Library By exhaustively deleting several components from the detailed chemical reaction model, and based on the relative error between the simplified reaction model and the detailed reaction model in predicting the simplified target value, the simplified reaction model is screened to obtain a reliable simplified reaction model library, which is then converted into a binary matrix to identify the component composition of each model. Step 2: Analysis of component coupling relationships and importance Using the relative error of the simplified model as the eigenvalue of the matrix constructed in step 1, and combining it with the Gaussian distribution and conditional probability formula, the probability of component A's presence in the reaction model can be obtained. Based on the state function curve, and using the error corresponding to a 99.99% probability of component presence as a basis, the importance of different components can be determined. Simultaneously, a virtual component is introduced. I By evaluating virtual components I The importance of clarifying virtual components I The strength of the coupling relationship among the components; Step 3: Simplify Target Definition Using the method in step 2, the detailed reaction model was simplified for three aspects: different pressures, different equivalence ratios, and different simplification targets (flame delay and component concentration) under wide temperature conditions. By comparing the structure of the simplified reaction model under different conditions, the influence of pressure, equivalence ratio, and simplification target on the structure of the simplified reaction model was summarized, and the simplified conditions and simplification targets were clarified. Step 4: Final Simplified Reaction Model Construction Using the simplified operating conditions and simplification objectives determined in step 3, the importance of the components in the detailed reaction model is ranked using the method in step 2; Subsequently, based on this sorting, the components are deleted one by one until the predicted value of the simplified reaction model under any operating condition exceeds the prediction uncertainty space of the detailed reaction model, thus obtaining the initial simplified model. Finally, a multi-objective genetic algorithm is used to optimize the reaction rate constant in the fuel sub-model within the uncertainty range to obtain the final simplified reaction model.

2. The novel method for identifying component coupling relationships in a chemical reaction model according to claim 1, characterized in that, Two to five components are randomly removed from the detailed reaction model each time using an exhaustive method.

3. A novel method for identifying component coupling relationships in a chemical reaction model according to claim 1, characterized in that, In the simplified reaction model, the retained components are set to "1", and the deleted components are set to "0". The total number of components in the detailed chemical reaction model is... m The total number of simplified reaction models in the simplified reaction model library is n， The simplified reaction model library is then transformed into a m × n A 3D matrix.

4. A novel method for identifying component coupling relationships in a chemical reaction model according to claim 1, characterized in that, Combining the Gaussian distribution and conditional probability formula, based on the components A Support S ( A The expected value of a normal distribution σ Components A The variance of the relative error during removal is a normal distribution. μ 1. Components A The variance of the relative error normal distribution when it exists μ 2; Components available A The existence probability function in the reaction model is formulated as follows: 。 5. A novel method for identifying component coupling relationships in a chemical reaction model according to claim 1, characterized in that, The virtual components I Used to determine the combination of research components, if I If all components are present, then I The state is "1" otherwise "0"; through virtual components I status Y I The first in the combination j The state of each component Y j The number of components in the combination J， Determine virtual components I The formula is: ； By evaluating virtual components I The importance of virtual components can be understood to obtain the coupling relationship between different components; virtual components I The more important, the more virtual components I The stronger the coupling relationship between the components contained therein.

6. A novel method for identifying component coupling relationships in a chemical reaction model according to claim 1, characterized in that, The detailed chemical reaction model is as follows: α -Methylnaphthalene reaction model.

Images