Engine transient performance optimization calibration method based on response curved surface design

Abstract

The invention is suitable for the technical field of engine performance calibration, and provides an engine transient performance optimization calibration method based on response curved surface design, comprising the following steps: setting decision variables and target variables to construct an experimental matrix; performing experimental test according to the test matrix; fitting and evaluating a quadratic model; performing simplified selection on the fitting model; performing prediction optimization calculation and experimental verification; and performing all-condition Map calibration according to the rotating speed range. The method can be practically applied to transient performance optimization in the research and development test process of the engine, the performance of the engine in the real vehicle running process can be optimized, the validity of test data is visually displayed through a mathematical model, and excessive test resources are not needed.

Description

Technical Field

[0001] The present invention belongs to the technical field of engine performance calibration, and particularly relates to an engine transient performance optimization calibration method based on response surface design. Background Art

[0002] The transient conditions of vehicle engines are increasingly regarded as the key to understanding engine performance and emissions. During the testing of environmental protection regulations and the actual operation of vehicles, most conditions are transient conditions, that is, real-time changing conditions. For example, in the case of acceleration, deceleration, or load change, due to the delay of intake response, it will significantly affect the engine's performance response and the formation of emissions such as soot and nitrogen oxides (NO x ) etc. A major area of concern in the study of diesel engine transient conditions is the transition of performance from steady-state conditions to transient conditions, especially in terms of fuel economy and emission control. At the same time, various test results show that there is a certain relationship between the boundary parameters of transient conditions, such as injection timing and injection pressure, and the change of emission concentrations by affecting combustion parameters. All parameters during the actual operation of the engine are based on the calibrated Map. Therefore, the actual performance of the engine has a direct relationship with the calibrated Map. For the calibration of traditional engine Maps, it is carried out under steady-state conditions, that is, all input parameters are fixed parameters, and then the optimal points of each parameter are found to obtain a set of optimal Maps under steady-state conditions. However, the actual vehicle operation is often with variable parameters and variable conditions, and it will often lead to significant deterioration of fuel consumption and emissions under severe conditions. Moreover, the calibration of such traditional Maps requires testing all working condition points, which is time-consuming and laborious.

[0003] The response surface method is a set of statistical and mathematical techniques used to develop, improve, and optimize processes, which can significantly reduce the number of experiments required to evaluate variable combinations and create a prediction model that can effectively capture the interactions between variables. In the context of diesel engines, using this method can explore the relationships between input factors, such as injection timing and injection pressure, and engine performance indicators. Therefore, combined with the defined transient condition calibration parameters, a set of optimized Maps based on transient condition calibration can be formulated, so as to be better applied to the actual operation of the engine. Summary of the Invention

[0004] The purpose of the embodiments of the present invention is to provide an engine transient performance optimization calibration method based on response surface design, aiming to solve the problems proposed in the above background art.

[0005] The embodiments of the present invention are implemented as follows. An engine transient performance optimization calibration method based on response surface design includes the following steps: Step 1: Set decision variables and target variables to construct an experimental matrix; Step 2: Conduct experimental tests according to the test matrix; Step 3: Fit and evaluate the quadratic model; Step 4: Simplify and select the fitted model; Step 5: Conduct predictive optimization calculations and experimental verification; Step 6: Calibrate the full operating condition Map within the speed range.

[0006] Further technical solution, Step 1 includes the following specific steps: The engine development user determines the input parameters and output parameters of the engine according to the transient operating conditions to be studied, and defines them as decision variables and target variables respectively.

[0007] The experimental matrix is set using orthogonal design. Assuming there are three parameters A, B, and C, the basis for the experimental matrix design is: ; ; ; ; Among them, c is the center point, that is, the point where all parameter levels are 0. Generally, the center point is repeated 3 - 5 times; for k decision variables, the number of experiments n is: ; Design the experimental plan according to the above rules, randomly shuffle its test order, and construct the experimental matrix.

[0008] Further technical solution, the decision variables are 3 - 6 parameters, which are divided into 3 level hierarchies according to the research range by the Box - Behnken method: -1, 0, and 1; the research decision variables include the engine speed and the load rate, providing a basis for quickly formulating the transient Map. The upper and lower limits of the engine speed are set according to the research speed range, and the load rate is the loading rate at which the engine loads the load from 0% to 100% at a constant speed, and is set by the torque per second, the load - loading ratio per second, or the process loading time.

[0009] Further technical solution, for the target variables, the following defined parameters are given: Take the effective fuel consumption rate and indicated thermal efficiency when the load reaches 50% during the transient process, and their variable coefficients under the corresponding steady - state operating conditions as performance indicators; take the time when the loaded load reaches 99% of the load as the stabilization time, and take the process mean emission concentrations of CO, NO x , Soot as emission indicators; Variable coefficient of effective fuel consumption rate The calculation formula is as follows: ; Among them, is the effective fuel consumption rate at 50% load under transient conditions, unit: g / kwh; is the effective fuel consumption rate at 50% load under the corresponding steady-state conditions, unit: g / kwh; Indicated thermal efficiency variable coefficient: ; CO process mean emission concentration: ; NO x Process mean emission concentration: ; Soot process mean emission concentration: ; Among them, is the real-time emission concentration value of CO, NOx, and Soot; is the time at 99% load; is the start loading time.

[0010] For a further technical solution, in the said step 2, tests are carried out according to the experimental matrix, and the recorded target variables are corresponded to the decision variables in the test matrix; rearrangement is made in the order of the change of the decision variables for fitting preparation.

[0011] For a further technical solution, the said step 3 includes the following specific steps: The general form of the quadratic model is: ; Among them, and are the parameters to be estimated, and are the i and j th decision variables, is the random error; written in matrix form: ; Among them, y is the target vector, X is the design matrix, , is the parameter vector, is the error vector, satisfying ; Minimize the sum of squared residuals by least squares estimation: ; Solve for the parameter estimates: For Take the derivative and set it to 0 to obtain the normal equations: ; ; If is invertible, the parameter estimates are: ; Then the quadratic model for each target variable is: ; Model evaluation metric calculation: Decompose and calculate the sum of squares of the data: ; Among them, is the total sum of squares, which is the fluctuation of the response value around the mean; is the regression sum of squares, which is the difference between the fitted value and the mean; is the residual sum of squares, which is the deviation between the predicted value and the actual value; The residual standard deviation is: ; Then the coefficient of determination : ; To avoid overfitting, penalize the insignificant variables in the model and introduce the adjusted coefficient value : ; Then the predicted coefficient value is: ; ; Among them, is the predicted residual sum of squares, is the predicted value refitted after removing the i th sample; Conduct the following test: Null hypothesis: The model parameters are not significant; Alternative hypothesis: The model parameters are significant; Calculate the value: ; For the calculated Obtain the corresponding P-value by looking up the value in the table. If the P-value is less than the significance level, the null hypothesis can be rejected; Introduce the coefficient of variation value to quantify the magnitude of the prediction error: ; Introduce the adaptation accuracy , which measures the balance between the prediction ability and complexity of the model: ; Summarize the parameter evaluation of the fitting for each target variable. If the evaluation parameters show good results, proceed to the next calculation. Otherwise, check whether the data is incorrect. If the data is incorrect, perform a correction test. If the data is correct and the performance evaluation is poor, it indicates that there is no correlation between the decision variable and the target variable.

[0012] Further technical solution, Above 0.8 indicates a good fitting model, If the value does not exceed 10%, it is considered that the model dispersion is within the acceptable range; If the value is greater than 4, the model is considered ideal.

[0013] Further technical solution, step 4 includes the following steps: Follow the Occam's razor principle, that is, do not add items unnecessarily, to simplify the model to avoid overfitting; eliminate the insignificant terms of the model to obtain a simplified model. By comparing the above evaluation parameters, preferentially select the simplified model on the premise that the performance does not deteriorate. Otherwise, keep the original model unchanged.

[0014] Further technical solution, step 5 includes the following steps: For the target variable to be optimized, transform its fitting model into the degree of expectation Perform multi-objective optimization; for the single-objective expected function, define the optimization objective and constraints. When maximizing the objective: ; When minimizing the objective: ; where s is the shape parameter. When s = 1, the expected function changes linearly; when s > 1, the expected function is a convex function; when s < 1, the expected function is a concave function. The overall expected function is: ; where, is the importance degree, which is set according to the requirements for the importance degree of the target variable; Use the gradient descent method to maximize : Set an initial point , which is the value of the decision variable at the start of iteration, and is denoted as during the iteration process. Randomly select or specify any point, and set the learning rate , the maximum number of iterations, and set the convergence threshold . When the parameter change is less than the threshold, it is considered to have converged, and the iteration is stopped; During the iteration process, calculate the overall expected function of the objective variable for each iteration point and calculate the gradient . For each decision variable , calculate its partial derivative: ; Update the variable: ; When , stop the iteration; List the operating condition points under the maximum overall expected function, that is, the optimal operating condition points under the model prediction; perform according to the decision variables of the optimal operating condition points or other predicted operating condition points.

[0015] Further technical solution. In step 6, for the formulation of the transient operating condition optimization Map, perform full operating condition optimization at a fixed speed and a fixed load rate; when performing optimization, set the importance according to the required target performance, and at the same time set the engine speed as a fixed speed starting from the lower limit of the research range, and start optimizing and predicting with a variable load rate; then gradually increase the speed at intervals of 100 r / min, and the engine load rate gradually changes according to the requirements, and perform optimization and prediction to establish the transient operating condition optimization Map.

[0016] An engine transient performance optimization and calibration method based on response surface design provided by an embodiment of the present invention can be actually applied to the performance calibration in the engine R & D test process, can optimize the performance of the engine during actual vehicle operation, and visually shows the effectiveness of the test data with a mathematical model, without excessive test resources. Brief Description of the Drawings

[0017] Figure 1 is a flowchart of an engine transient performance optimization and calibration method based on response surface design provided by an embodiment of the present invention; Figure 2 is a map lookup diagram for actual application operating conditions. Detailed Embodiment

[0018] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0019] The following describes in detail the specific implementation of the present invention with reference to specific embodiments.

[0020] As Figure 1 shown, a method for optimizing and calibrating the transient performance of an engine based on response surface design provided by an embodiment of the present invention includes the following steps: Step 1: Set decision variables and target variables to construct an experimental matrix; Step 2: Conduct experimental tests according to the test matrix; Step 3: Fit and evaluate a quadratic model; Step 4: Simplify and select the fitted model; Step 5: Conduct predictive optimization calculations and experimental verifications; Step 6: Perform full-condition Map calibration within the speed range.

[0021] As a preferred embodiment of the present invention, the specific steps of the said Step 1 include: The engine development user determines the input parameters and output parameters of the engine according to the transient conditions to be studied for the engine, and defines them as decision variables and target variables respectively.

[0022] The experimental matrix is set using orthogonal design. For example, in the case of three parameters A, B, and C, the basis for the experimental matrix design is: ; ; ; ; where c is the center point, that is, the point where all parameter levels are 0, and generally the center point is repeated 3 - 5 times; for k decision variables, the number of experiments n is: ; According to the above rules, an experimental plan is designed, and its test order is randomly shuffled to construct an experimental matrix.

[0023] In the embodiments of the present invention, the decision variables are generally 3 - 6 parameters, which are divided into 3 hierarchical levels according to the research scope by the Box - Behnken method: -1, 0, and 1. It is recommended to include the engine speed and the load rate in the research decision variables to provide a basis for quickly formulating the transient Map. The upper and lower limits of the engine speed are set according to the research speed range, generally within the range of 800 r / min - 2200 r / min. The load rate is the loading rate at which the engine loads the load from 0% to 100% at a constant speed, and is set in terms of the torque loaded per second (Nm / s), the load loading ratio per second (% / s), or the process loading time (s).

[0024] For the target variables, the following defined parameters are given, and users can expand according to actual needs: Take the effective fuel consumption rate and the indicated thermal efficiency when the load reaches 50% during the transient process, and their variable coefficients under the corresponding steady - state conditions as performance indicators. Take the time when the loaded load reaches 99% of the load as the stabilization time, and take the process mean emission concentrations of CO, NO x , and Soot during the time to reach the stabilization time as emission indicators.

[0025] Variable coefficient of effective fuel consumption rate (%) is calculated as follows: ; where is the effective fuel consumption rate at 50% load under transient conditions, unit: g / kwh; is the effective fuel consumption rate at 50% load under the corresponding steady - state conditions, unit: g / kwh; Variable coefficient of indicated thermal efficiency (%): ; CO process mean emission concentration (unit: ppm): ; NO x process mean emission concentration (unit: ppm): ; Soot process mean emission concentration (unit: 1012n / cc): ; where is the real - time emission concentration value of CO, NOx, and Soot; is the time at 99% load, unit: s; is the start loading time, unit: s.

[0026] As a preferred embodiment of the present invention, in the step 2, tests are carried out according to the experimental matrix, and the obtained target variables are corresponded to the decision variables in the test matrix; rearrangement is made in the order of change of the decision variables to prepare for fitting.

[0027] As a preferred embodiment of the present invention, the step 3 includes the following specific steps: The quadratic model can describe the non-linear relationship between the target variable and multiple decision variables, and its general form is: ; Among them, and are the parameters to be estimated, and are the i and j th decision variables, is the random error; written in matrix form: ; Among them, y is the target vector, X is the design matrix, , is the parameter vector, is the error vector, satisfying .

[0028] Minimize the sum of squared residuals with least squares estimation: ; Solve for the parameter estimation: Take the derivative of and set the derivative to 0 to obtain the normal equation: ; ; If is invertible, the parameter estimation value is: ; Then for each target variable, its quadratic model is: ; Model evaluation index calculation: Decompose and calculate the sum of squares of the data: ; Among them, is the total sum of squares, which is the fluctuation of the response value around the mean; is the regression sum of squares, which is the difference between the fitted value and the mean; is the sum of squared residuals, which represents the deviation between the predicted value and the actual value. The residual standard deviation is then: ; Then the coefficient of determination is: ; To avoid overfitting, a penalty is imposed on the insignificant variables in the model, and an adjusted coefficient value is introduced: ; Then the predicted coefficient value is: ; ; where, is the predicted sum of squared residuals, is the predicted value re-fitted after removing the i th sample.

[0029] Usually, the following hypotheses are tested: Null hypothesis (H0): The model parameters are not significant.

[0030] Alternative hypothesis (H1): The model parameters are significant.

[0031] Calculate the value: ; Look up the corresponding P-value in the table for the calculated value. If the P-value is less than the significance level (usually 0.05), the null hypothesis can be rejected.

[0032] The coefficient of variation value (%) is introduced to quantify the magnitude of the prediction error: ; The adaptation accuracy is introduced to measure the balance between the prediction ability and complexity of the model: ; For each target variable, the parameter evaluation of the fitting is summarized. If the evaluation parameters show good results, the next calculation is carried out. Otherwise, it is necessary to check whether the data is incorrect. If the data is incorrect, a correction test is required. If the data is correct and the performance evaluation is poor, it means that the decision variable has no correlation with the target variable.

[0033] In the embodiment of the present invention, Above 0.8 indicates a good fitting model, If the value does not exceed 10%, it is considered that the model dispersion is within the acceptable range. The larger the value, the better the predictive ability of the model under the given complexity. Generally, if it is greater than 4, the model is considered ideal, which indicates that the model has a good signal-to-noise ratio and can effectively predict the response variable.

[0034] As a preferred embodiment of the present invention, step 4 includes the following steps: Follow the Occam's razor principle, that is, do not add items unnecessarily, to simplify the model to avoid overfitting. Eliminate the insignificant terms of the model to obtain a simplified model. By comparing the above evaluation parameters, preferentially select the simplified model on the premise that the performance does not deteriorate, otherwise keep the original model unchanged.

[0035] As a preferred embodiment of the present invention, step 5 includes the following steps: For the target variable to be optimized, convert its fitting model into a desirability Perform multi-objective optimization. For a single-objective desirability function, define the optimization objective and constraints. When maximizing the objective: ; When minimizing the objective: ; where s is the shape parameter. When s = 1, the desirability function changes linearly; when s > 1, the desirability function is a convex function; when s < 1, the desirability function is a concave function. The overall desirability function is: ; where, is the importance degree, and the user can set the importance degree of the target variable according to the requirements.

[0036] Use the gradient descent method to maximize : Set an initial point , which is the value of the decision variable at the start of iteration and is denoted as during the iteration process. It can be randomly selected or specified at any point. Set the learning rate , such as 0.01, the maximum number of iterations, such as 1000, and set the convergence threshold , such as 1e-6. When the parameter change is less than the threshold, it is considered to have converged and the iteration stops.

[0037] During the iteration process, calculate the overall desirability function of the target variable for each iteration point and calculate the gradient , and for each decision variable , calculate its partial derivative: ; Update variable: ; When , stop the iteration.

[0038] List the operating points under the maximum overall expectation function, that is, the optimal operating points under model prediction. Proceed according to the decision variables of the optimal operating points or other predicted operating points.

[0039] As a preferred embodiment of the present invention, in step 6, for the formulation of the transient operating condition optimization Map, full operating condition optimization is carried out at a constant speed and a constant load rate. When performing the optimization, set the importance according to the required target performance. For example, the importance of engine thermal efficiency and fuel consumption is set to 5, and other performance parameters are set to 3. At the same time, set the engine speed as a fixed speed starting from the lower limit of the research range, such as 800 r / min, and start optimizing and predicting with a variable load rate. Then gradually increase the speed at intervals of 100 r / min, and the engine load rate gradually changes according to the requirements to optimize and predict to establish a transient operating condition optimization Map.

[0040] The following provides a specific embodiment to verify the effectiveness of the method: A certain 6-cylinder diesel engine is carried out under three decision variables: speed 900 - 1300 r / min, load time 3 - 7 s, and fuel injection advance angle -2 - 2°. The experimental design matrix, model evaluation parameters, and significance analysis are shown in Tables 1 - 3 below.

[0041] Table 1 Experimental design matrix

[0042] Table 2 Model evaluation parameter table

[0043] Table 3 Simplified model significance analysis table

[0044] In Table 3, all the fitting models of the target variables are significant. Although the P value of the AC interaction term of Co BSFC is higher than 0.05 but the excess is not large, it is retained to ensure the structural integrity. Removing the three interaction terms of mE CO will affect the model performance and is retained. The interaction coefficients of mE NOx and mE Soot are all significant. In addition, the linear and second-order coefficients of Co BSFC , Co BTE and mE Soot are significant. For Co BSFC , R2 The value is 0.980, which is close to 1, indicating the accuracy and sufficiency of the model. The value (0.965) and the value (0.918) are both large, and the model has good predictive ability. Co BSFC The value of the adaptation accuracy is 24.3, and the coefficient of variation is 6.05. Judging from the combination of residuals, standard deviation and mean, the model is effective. Similarly, Co BTE , mE CO , mE NOx and mE Soot The evaluation of the model parameters (Table 2) shows that the values are all above 0.85, and the adaptation accuracy is far greater than 4, indicating that each model has appropriate accuracy, and from the perspectives of residuals, standard deviation, mean and coefficient of variation, the degree of dispersion is within an acceptable range. The obtained simplified and screened regression equation is: ; ; ; ; ; The verification analysis of the model is shown in Table 4. Set Co BSFC and Co ITE The importance is 5, and each emission index is 3 to formulate the transient performance optimization Map as shown in Table 5. Figure 2 a is the curve of a certain actual operating condition, Figure 2 while b is the curve for reading the injection advance angle after applying the transient performance optimization Map after converting the torque change into the loading time.

[0045] Table 4 Decision variable range

[0046] Table 5 Formulating the transient optimization Map

[0047] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. An engine transient performance optimization calibration method based on response surface design, characterized in that, The following steps are involved: Step 1: Set the decision variables and target variables to construct the experimental matrix; Step 2: Conduct experimental testing according to the test matrix; Step 3: Fit and evaluate the quadratic model; Step 4: Simplify and select the fitting model; Step 5: Perform prediction optimization calculation and experimental verification; Step 6: Perform full-condition map calibration over the speed range.

2. The engine transient performance optimization calibration method based on response surface design according to claim 1, wherein The step 1 comprises the following specific steps: The engine development user determines the input parameters and output parameters of the engine according to the transient operating conditions to be studied, and defines them as decision variables and target variables respectively; The experimental matrix setting adopts orthogonal design. Assuming that there are three parameters A, B, and C, the experimental matrix design is based on: ； ； ； ； Among them, c is the center point, that is, the point where all parameter levels are 0, and the center point is repeated 3 - 5 times; for k decision variables, the number of experiments n is: ； According to the above rules, an experimental plan is designed, the test order is randomly disrupted, and an experimental matrix is ​​constructed.

3. The engine transient performance optimization calibration method based on response surface design according to claim 2, characterized in that, The decision variables are 3-6 parameters, which are divided into 3 levels according to the research scope using the Box-Behnken method: -1, 0 and 1; The research decision variables include engine speed and loading rate to provide a basis for quickly formulating transient maps. The engine speed is set with upper and lower limits based on the research speed range. The loading rate is set based on the loading rate at which the engine loads the load from 0% to 100% at a constant speed, in loading torque per second, load ratio per second, or process loading time.

4. The engine transient performance optimization calibration method based on response surface design according to claim 2, wherein For the target variable, the following definition parameters are given: Taking the effective fuel consumption rate and indicated thermal efficiency at 50% load during the transient process and their variable coefficients under the corresponding steady-state conditions as performance indicators; taking the time when the loading reaches 99% load as the stabilization time, and taking the process mean emission concentrations of CO, NO x , and Soot within the stabilization time as emission indicators; Variable coefficient of effective fuel consumption rate The calculation formula is as follows: ； Among them, is the effective fuel consumption rate at 50% load under transient conditions, unit: g / kwh; is the effective fuel consumption rate at 50% load under the corresponding steady-state conditions, unit: g / kwh; Indicated thermal efficiency variable coefficient: ； CO process average emission concentration: ； NO x Process mean emission concentration: ； Soot process mean emission concentration: ； Among them, are the real-time emission concentration values of CO, NOx, and Soot; is the time at 99% load; is the start loading time.

5. The engine transient performance optimization calibration method based on response surface design according to claim 1, characterized in that, In step 2, the test is performed according to the experimental matrix, and the recorded target variables are matched with the decision variables in the test matrix; the decision variables are rearranged in the order of change to prepare for fitting.

6. The engine transient performance optimization calibration method based on response surface design according to claim 4, characterized in that The step 3 comprises the following specific steps: The general form of the quadratic model is: ； Among them, and are parameters to be estimated, and are the i and j th decision variables, is the random error; written in matrix form: ； wherein, y is 's target vector, X is 's design matrix, , is 's parameter vector, is 's error vector, satisfying ; Minimize the residual sum of squares using least squares estimation: ； Solve for parameter estimates: For Take the derivative and set it equal to 0 to obtain the normal equation: ； ； If is reversible, the parameter estimate is: ； Then for each target variable, the quadratic model is: ； Model evaluation index calculation: Decompose and calculate the sum of squares of the data: ； Among them, is the total sum of squares, which is the fluctuation of the response value around the mean; is the regression sum of squares, which is the difference between the fitted value and the mean; is the residual sum of squares, which is the deviation between the predicted value and the actual value; the residual standard deviation is: ； Then the coefficient of determination : ； To avoid overfitting, penalize the insignificant variables in the model and introduce the adjustment coefficient value : ； Then the prediction coefficient value is as follows: ； ； Among them, is the sum of squared prediction residuals, is the predicted value refitted after removing the i th sample. Verify the following assumptions as follows: Null hypothesis: the model parameters are not significant; Alternative hypothesis: The model parameters are significant; Calculation Value: ； For the calculated value, look up the corresponding P-value in the table. If the P-value is less than the significance level, the null hypothesis can be rejected; Introduce the coefficient of variation value to quantify the magnitude of the prediction error: ； Introduce adaptation accuracy , which measures the balance between the prediction ability and complexity of the model: ； The parameter evaluation of the fit of each target variable is summarized. If the evaluation parameters show good results, the next step of calculation is carried out. Otherwise, the data needs to be checked for errors. If the data is incorrect, a correction test is performed. If the data is correct but the performance evaluation is poor, it means that the decision variable has no correlation with the target variable.

7. The engine transient performance optimization calibration method based on response surface design according to claim 6, characterized in that, Above 0.8 indicates a good fitting model, If the value does not exceed 10%, the model dispersion is considered within the acceptable range; If the value is greater than 4, the model is considered ideal.

8. The engine transient performance optimization calibration method based on response surface design according to claim 6, characterized in that The step 4 comprises the following steps: Follow the principle of Occam's razor, that is, do not add items unless necessary, to simplify the model to avoid overfitting; remove the insignificant items in the model to obtain a simplified model. By comparing the above evaluation parameters, the simplified model is selected first under the premise that the performance is not deteriorated, otherwise the original model remains unchanged.

9. The engine transient performance optimization calibration method based on response surface design according to claim 8, characterized in that The step 5 comprises the following steps: For the target variable to be optimized, convert its fitting model into an expected degree Perform multi-objective optimization; for a single-objective expected function, define the optimization objective and constraints. When maximizing the objective: ； When minimizing the objective: ； where s is the shape parameter. When s = 1, the expected function changes linearly; when s > 1, the expected function is a convex function; when s < 1, the expected function is a concave function. The overall expected function is as follows: ； Among them, is the importance degree, which is set according to the requirements for the importance degree of the target variable; Maximize using gradient descent : Set an initial point , which is the value of the decision variable at the start of iteration and is denoted as during the iteration process. Randomly select or specify any point, and set the learning rate , the maximum number of iterations, and set the convergence threshold . When the change in the parameter is less than the threshold, it is considered to have converged and the iteration stops; During the iteration process, calculate the overall expected function of the target variable for each iteration point and calculate the gradient , for each decision variable , calculate its partial derivative: ； Update variables: ； When stop iterating; List the operating points under the maximum overall expected function, that is, the optimal operating points under the model prediction; proceed according to the decision variables of the optimal operating point or other predicted operating points.

10. The engine transient performance optimization calibration method based on response surface design according to claim 9, characterized in that, In step 6, for the formulation of the transient condition optimization Map, full-condition optimization is carried out at a fixed speed and a fixed load rate; when optimizing, the importance is set according to the required target performance, and at the same time, the engine speed is set as a fixed speed starting from the lower limit of the research range, and the variable load rate starts to optimize and predict; Then the speed is gradually increased at intervals of 100 r / min, and the engine load rate gradually changes according to the requirements, and optimization and prediction are carried out to establish the transient condition optimization Map.

Images