Calibration method for improving precision of small and medium load air charging model

Abstract

The application discloses a calibration method for improving the precision of a small and medium load charging model, comprising the following steps: carrying out a charging model scanning test on an engine test bench to obtain original charging model scanning data; cleaning and arranging the original scanning data to obtain qualified data; classifying the cleaned and arranged data, and dividing the data into small load, medium load and large load according to a predetermined threshold; weighting the data points of different loads by different multiples; performing fitting calculation by using the weighted data, adjusting the charging model parameters through multiple iterations, continuously reducing the RMSE, and finally converging to obtain a group of charging model parameters; importing the finally obtained charging model parameters into an ECU, carrying out verification work of the charging model on the engine test bench, and judging whether the main charging model precision of the engine under the universal characteristic condition meets the requirements. The application realizes great improvement of the precision of the small and medium load charging model.

Description

Technical Field

[0001] This invention belongs to the field of engine calibration technology, and particularly relates to a calibration method for improving the accuracy of a small-to-medium load charging model. Background Technology

[0002] The process of matching an engine with an electronic fuel injection system requires precise measurement of engine speed and intake air volume. This allows for precise control of the fuel (gas) injection quantity and ignition advance angle, as well as the air-fuel ratio, to achieve optimal engine power, fuel economy, and emissions. While the crankshaft position sensor can easily measure engine speed, accurately calculating the intake air volume under various engine operating conditions is an extremely complex calibration process—the process of establishing a charging model. Establishing the charging model is the primary task in calibrating the engine's electronic fuel injection system. Subsequent calibration work is based on this established charging model. The accuracy of the charging model significantly affects the overall calibration quality of the electronic fuel injection system; therefore, it is necessary to evaluate the accuracy of the charging model.

[0003] Existing technical solutions:

[0004] The patent application CN110926821A, entitled "An Evaluation Method for the Accuracy of an Engine Inflation Model", mentions that the accuracy of the engine inflation model can be evaluated by calculating the percentage difference between the engine intake volume calculated by the ECU under various operating conditions and the engine intake volume measured by existing equipment.

[0005] The existing solution is to improve the accuracy of the charging model by weighting the data of universal characteristic conditions (i.e., the final operating conditions of the engine: fixed combination of speed / load / valve timing). Although this solution has a certain improvement effect on the overall accuracy of the engine charging model, the improvement on the accuracy of the charging model under medium and small loads is extremely insignificant. Summary of the Invention

[0006] Based on the problems and defects of existing technologies, this invention proposes a calibration method to improve the accuracy of small and medium load inflation models. This method can greatly improve the accuracy of small and medium load inflation models, thereby shortening the engine calibration cycle, improving the engine's operating performance under small and medium loads, and enhancing product competitiveness.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A calibration method for improving the accuracy of air-filled models under medium and low loads, comprising:

[0009] Step 1: Conduct an inflation model scanning test on the engine bench to obtain the original inflation model scanning data;

[0010] Step 2: Clean and organize the raw scanned data to obtain data that meets the requirements;

[0011] Step 3: Classify the cleaned and organized data, dividing it into low-load, medium-load, and high-load categories according to predetermined thresholds;

[0012] Step 4: Weight the data points with different loads by different factors;

[0013] Step 5: Use the weighted data for fitting calculation. After multiple iterations, adjust the inflatable model parameters to continuously reduce the RMSE, and finally converge to obtain a set of inflatable model parameters.

[0014] Step 6: Import the final inflation model parameters into the ECU and verify the inflation model on the engine bench to determine whether the accuracy of the engine's main inflation model meets the requirements under universal characteristic conditions. If it does not meet the requirements, repeat steps 3 to 6 until the requirements are met.

[0015] As a further aspect of the present invention, the scanning test in the first step involves scanning different valve timing combinations at different speeds and loads according to the test plan.

[0016] As a further aspect of the present invention, the data cleaning in the second step involves removing noisy data according to predetermined cleaning rules. If the cleaning does not provide sufficient data coverage for key operating conditions, the process returns to the first step to supplement the original data scan.

[0017] As a further aspect of the present invention, the third step involves classifying the cleaned and organized data, including:

[0018] The operating condition where the actual inflation efficiency is less than the load TS is marked as the low load operating condition, and the number of test points for the low load operating condition is calculated as a;

[0019] The operating condition where the actual inflation efficiency is between the load TS and the load TM is marked as the medium load operating condition, and the number of test points for the medium load operating condition is calculated as b;

[0020] The operating condition where the actual inflation efficiency is greater than the load TM is marked as the high load operating condition, and the number of test points for the high load operating condition is calculated as c.

[0021] As a further aspect of the present invention, in the fourth step, data points with different loads are weighted by different multiples, including:

[0022] The test points under low load conditions are weighted by x times, and the number of test points under low load conditions participating in the fitting is calculated as ax;

[0023] The test points under medium load conditions are weighted by y times, and the number of medium load condition test points participating in the fitting is calculated as bx;

[0024] The test points under high load conditions are weighted by 1, and the number of test points under high load conditions participating in the fitting is calculated as c.

[0025] The total number of data points involved in the fitting calculation is calculated as n = ax + by + c.

[0026] As a further aspect of the present invention, in the fifth step, the fitting calculation includes:

[0027] S1: Perform fitting calculations on the weighted n data points;

[0028] S2: Set the iteration number g = 1;

[0029] S3: Begin the g-th adjustment of model parameters;

[0030] S4: Output model parameters are denoted as PARA. g The corresponding root mean square error (RMSE) g ;

[0031] S5: Begin the (g+1)th adjustment of model parameters;

[0032] S6: The output model parameters are denoted as PARAg+1, and the corresponding root mean square error RMSEg+1;

[0033] S7: Decision | RMSE g+1 -RMSE g |≥k;

[0034] If so, record g = g + 1 and jump back to S5;

[0035] If not, output the final model parameter RMSEg+1.

[0036] As a further aspect of the present invention, in the sixth step, when determining whether the accuracy of the inflatable model meets the expected requirements, if it does not, adjust TS, TM, x, and y and repeat steps three to six until the requirements are met.

[0037] By adopting the above technical solutions, this invention can provide the following beneficial effects in the process of calibrating the accuracy of engine inflation models:

[0038] (1) This invention addresses the acceptance criteria for inflation efficiency by rationally applying weighting to achieve a direct and effective improvement in the accuracy of inflation models under small and medium loads.

[0039] (2) This invention can be used not only for the accuracy optimization of inflatable models under medium and small loads, but also for other working conditions, thereby achieving direct improvement of the accuracy of inflatable models in that area.

[0040] (3) This invention can be used not only for the optimization of the accuracy of the engine charging model, but also for the optimization of other models with RMSE (root mean square error) as the optimization objective. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 A flowchart illustrating the existing method for calibrating inflatable models.

[0043] Figure 2 This is a schematic diagram of the process of the inflatable model scanning test, the first step of the calibration method provided in this embodiment of the invention.

[0044] Figure 3 This is a schematic diagram of the scanning data cleaning and processing process in the second step of the calibration method provided in this embodiment of the invention.

[0045] Figure 4 This is a schematic diagram of the data classification flow in the third step of the calibration method provided in this embodiment of the invention.

[0046] Figure 5 This is a flowchart illustrating the data weighting process in the fourth step of the calibration method provided in this embodiment of the invention.

[0047] Figure 6 This is a schematic diagram of the fitting calculation process in the fifth step of the calibration method provided in this embodiment of the invention.

[0048] Figure 7 This is a schematic diagram of the sixth step of the calibration method provided in this embodiment of the invention: verification of the parameters of the inflatable model. Detailed Implementation

[0049] The present invention aims to propose a calibration method to improve the accuracy of the low-load charging model of an engine.

[0050] The charging efficiency characterizes the amount of air intake in an engine and is an important basis for the control of engine fuel injection and ignition. Its model accuracy affects the engine's fuel consumption, emissions, and power performance. During the vehicle's life cycle, the engine operates under medium and low load conditions for a long time, so the accuracy of the charging model under medium and low load is very important.

[0051] See Figure 1As shown, the calibration of an engine inflation model typically involves the following steps: (1) conducting a scanning test on an engine bench to collect data; (2) cleaning and organizing the scanning data on a PC; (3) performing fitting calculations on the organized data on a PC to output inflation model parameters; and (4) importing the inflation model parameters into the engine controller and verifying the inflation model accuracy on an engine bench. As mentioned above, the inflation model parameters directly affect the accuracy of the inflation model; and the inflation model parameters are mainly affected by: the quality of the scanning test data, data cleaning and organization, and fitting calculations.

[0052] Because the quality of test data under low and medium loads is difficult to improve (engine combustion is relatively unstable under low loads, and the accuracy of measuring equipment is also relatively low) and the fitting algorithm is difficult to optimize, the accuracy of models under low and medium loads has always been difficult to improve. This invention proposes a calibration method to specifically improve the accuracy of low and medium load inflation models by weighting the scan data under low and medium loads. Its working logic is as follows:

[0053] Acceptance criteria for the accuracy of the engine charging model: (1) The relative error of the universal characteristic at 100% operating point is within ±10%; (2) The relative error of the universal characteristic at 90% operating point is within ±5%. The fitting objective of the engine charging model is to minimize the root mean square error (RMSE) of the operating points involved in the fitting.

[0054] Relative error of inflatable model

[0055] Root mean square error

[0056]

[0057] Comparing the two formulas above, it can be seen that the acceptance standard is for relative error, while the fitting target is for absolute error. When the relative error is constant, the absolute error under low-load conditions is much smaller than that under high-load conditions (for example, with a relative error requirement of ±5%, a low-load inflation efficiency of 20% corresponds to an absolute error of the model inflation efficiency within ±1% (±5% * 20%); while for a high-load inflation efficiency of 200%, the absolute error of the model inflation efficiency is required to be within ±5% * 200% = ±10%). In the fitting calculation, since all load conditions will ultimately use the same set of model parameters, the fitting calculation will prioritize optimizing the high-load conditions to minimize the RMSE (the RMSE gain from optimizing the conditions with large absolute error deviations is significantly greater than the gain from optimizing the conditions with small absolute error deviations).

[0058] As described above, this invention artificially increases the influence factor of small and medium load test data in the fitting calculation process by weighting the small and medium load test data (increasing the number of small and medium load test points), so that the fitted model parameters can be tilted towards small and medium loads, thereby achieving a significant improvement in the accuracy of the small and medium load inflation model while hardly reducing the accuracy of the high load inflation model.

[0059] This invention proposes a calibration method to improve the accuracy of inflatable models under medium and small loads, which mainly includes the following steps:

[0060] Step 1: Conduct a gas-filled model scanning test on an engine bench to obtain the original gas-filled model scanning data. The scanning test involves scanning different valve timing combinations at different speeds and loads according to the test plan.

[0061] Step 2: Clean and organize the original scanned data on the PC to obtain data that meets the requirements. Data cleaning involves removing noisy data according to established cleaning rules to avoid its impact on the subsequent fitting process. If the amount of cleaned data is too small or does not adequately cover key operating conditions, it is necessary to return to step 1 to supplement the original scanned data.

[0062] Step 3: Classify the cleaned and organized data on the PC, dividing the data into low-load, medium-load, and high-load categories according to predetermined thresholds, so that individual weighting can be performed later.

[0063] Step 4: On the PC, the data points with different loads are weighted by different factors, and the weighted data is used as the final data for subsequent fitting calculations.

[0064] Step 5: Use the above data on the PC to perform fitting calculations. After multiple iterations, adjust the inflatable model parameters to continuously reduce the RMSE, and finally converge to obtain a set of inflatable model parameters.

[0065] Step 6: Import the inflation model parameters obtained above into the ECU, and perform the inflation model verification work on the engine bench to determine whether the accuracy of the main inflation model of the engine meets the requirements under universal characteristic conditions. If it does not meet the requirements, repeat steps 3 to 6 until the requirements are met.

[0066] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0067] A calibration method for improving the accuracy of air-filled models under medium and low loads comprises the following steps:

[0068] (1) Step 1: Develop a main charge scan test plan (k valve timing groups, k can be defined, e.g., k=30); build an engine bench and start the main charge scan test: collect charge scan data under different speeds and load conditions for different valve timing combinations (combination 1, combination 2, ..., combination k); complete the main charge scan test and obtain the charge scan data, such as... Figure 2 As shown.

[0069] (2) Step 2: Clean and organize the experimental data obtained from Step 1. Remove unsuitable scan data according to established rules (r data cleaning rules, where r can be defined, e.g., r = 20). If necessary, repeat Step 1 to supplement the scan data. Figure 3 As shown, the established rules can be:

[0070] According to cleaning rule 1: if the air-fuel ratio is greater than the threshold, data that does not meet the requirements will be cleaned.

[0071] According to cleaning rule 2: if the combustion cycle fluctuation rate exceeds the threshold, the data that does not meet the requirements will be cleaned.

[0072] ...

[0073] According to the cleaning rule r: if the "standard deviation of fuel consumption rate / fuel consumption rate" exceeds the threshold, the data that does not meet the requirements will be cleaned.

[0074] If the amount of data after cleaning is too small or does not provide sufficient coverage for key operating conditions, it is necessary to return to Step 1 to supplement the original data scan.

[0075] (3) Step 3: Classify the scan data after Step 2: Assume that the condition where the actual inflation efficiency is less than the load TS is the low-load condition (the threshold TS can be defined, such as TS = 50%); the condition where the actual inflation efficiency is between the load TS and the load TM is the medium-load condition (the threshold TM can be defined, such as TM = 100%); the condition where the actual inflation efficiency is greater than the load TM is the high-load condition; calculate the number of test points for the low-load condition as a, the number of test points for the medium-load condition as b, and the number of test points for the high-load condition as c according to the above rules, such as... Figure 4 As shown.

[0076] (4) Step 4: Weight the low-load operating points obtained from Step 3 by a factor of x (the factor x can be defined, such as x = 5); weight the medium-load operating points by a factor of y (the factor y can be defined, such as y = 3); weight the high-load operating points by a factor of 1 (equivalent to no weighting); calculate all data involved in the fitting calculation as n = ax + by + c, such as... Figure 5 As shown.

[0077] (5) Step 5: Use the n data points weighted in Step 4 to perform fitting calculations, and adjust the parameters of the inflatable model through multiple iterations to achieve the minimum root mean square error (RMSE).

[0078] like Figure 6 As shown, the steps for fitting calculation are as follows:

[0079] S1: Perform fitting calculations on the weighted n data points;

[0080] S2: Set the iteration number g = 1;

[0081] S3: Begin the g-th adjustment of model parameters;

[0082] S4: Output model parameters are denoted as PARA. g The corresponding root mean square error (RMSE) g ;

[0083] S5: Begin the (g+1)th adjustment of model parameters;

[0084] S6: Output model parameters are denoted as PARA. g+1 The corresponding root mean square error (RMSE) g+1 ;

[0085] S7: Decision | RMSE g+1 -RMSE g |≥k;

[0086] If so, record g = g + 1 and jump back to S5;

[0087] If not, complete the fitting calculation and output the final model parameters RMSE. g+1 .

[0088] (6) Step 6: Import the inflation model parameters obtained from Step 5 into the ECU (engine controller) and test and verify them on the engine bench. Calculate the relative error diff of the inflation model at the universal characteristic operating point. rl % to determine whether the accuracy of the inflatable model meets the expected requirements. If not, adjust TS, TM, x, and y and repeat Step 3-6 until the requirements are met.

[0089] As mentioned above, Step 3 (classifying the data) and Step 4 (assigning different weights to different types of data) are the core steps of this invention.

[0090] The innovation of this invention lies in its rational application of weighting to the acceptance criteria for inflation efficiency, thereby achieving a direct and effective improvement in the accuracy of inflation models under small to medium loads. This innovative approach can not only be used for optimizing the accuracy of inflation models under small to medium loads, but also for other operating conditions, thus directly improving the accuracy of inflation models in those regions. Furthermore, this innovative approach can be used not only for optimizing the accuracy of engine inflation models, but also for optimizing other models that use RMSE (Root Mean Square Error) as the optimization objective.

[0091] This invention represents a technological breakthrough, enabling a significant improvement in the accuracy of medium and low load inflation models while maintaining almost no reduction in the accuracy of high-load inflation models. This, in turn, shortens the engine calibration cycle, improves engine performance under medium and low loads, and enhances product competitiveness.

[0092] Although embodiments of the present invention have been shown and described, these specific embodiments are merely explanations of the invention and are not intended to limit it. The specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. After reading this specification, those skilled in the art may make modifications, substitutions, and variations to the embodiments as needed without departing from the principles and spirit of the invention, but such modifications, substitutions, and variations are protected by patent law as long as they are within the scope of the claims of the present invention.

Claims

1. A calibration method for improving the accuracy of a small and medium load pneumatic model, characterized in that, include: Step 1: Conduct an inflation model scanning test on the engine bench to obtain the original inflation model scanning data; Step 2: Clean and organize the raw scanned data to obtain data that meets the requirements; Step 3: Classify the cleaned and organized data, dividing it into low-load, medium-load, and high-load categories according to predetermined thresholds; Step 4: Weight the data points with different loads by different factors; Step 5: Use the weighted data for fitting calculation. After multiple iterations, adjust the inflatable model parameters to continuously reduce the RMSE, and finally converge to obtain a set of inflatable model parameters. Step 6: Import the final inflation model parameters into the ECU and verify the inflation model on the engine bench to determine whether the accuracy of the engine's main inflation model meets the requirements under universal characteristic conditions. If it does not meet the requirements, repeat steps 3 to 6 until the requirements are met. The third step involves classifying the cleaned and organized data, including: The condition where the actual inflation efficiency is less than the load TS is marked as a low-load condition. The threshold TS can be defined, and the number of test points for the low-load condition is calculated as a. The operating condition where the actual inflation efficiency is between the load TS and the load TM is marked as the medium load operating condition. The threshold TM can be defined, and the number of test points for the medium load operating condition is calculated as b. The operating condition where the actual inflation efficiency is greater than the load TM is marked as the high load operating condition, and the number of test points for the high load operating condition is calculated as c. In the fourth step, data points with different loads are weighted by different factors, including: The test points under low load conditions are weighted by x times, and the number of test points under low load conditions participating in the fitting is calculated as ax; The test points under medium load conditions are weighted by y times, and the number of medium load condition test points participating in the fitting is calculated as by. The test points under high load conditions are weighted by 1, and the number of test points under high load conditions participating in the fitting is calculated as c. The total number of data points involved in the fitting calculation is calculated as n = ax + by + c.

2. The calibration method for improving the accuracy of a part-load air charge model according to claim 1, characterized in that, The first step of the scanning test involves scanning different valve timing combinations at different speeds and loads according to the test plan.

3. The calibration method for improving the accuracy of a part-load air charge model according to claim 1, characterized in that, The second step, data cleaning, involves removing noisy data according to established cleaning rules. If the cleaned data does not provide sufficient coverage of key operating conditions, the process returns to the first step to supplement the original data scan.

4. The calibration method for improving the accuracy of a part-load air charge model according to claim 1, characterized in that, In the fifth step, the fitting calculation includes: S1: Perform fitting calculations on the weighted n data points; S2: Set the iteration number g=1; S3: Begin the g-th adjustment of model parameters; S4: output the model parameters, denoted as PARA g , the corresponding root mean square error RMSE g ; S5: Begin the (g+1)th adjustment of model parameters; S6: output the model parameters, denoted as PARA g+1 The corresponding root mean square error RMSE g+1 ; S7: judging ; If so, then record g = g + 1 and jump back to S5; If not, output final model parameters RMSE g+1 .

5. The calibration method for improving the accuracy of a part-load air charge model according to claim 4, characterized in that, In step six, when determining whether the accuracy of the inflatable model meets the expected requirements, if it does not, adjust TS, TM, x, and y and repeat steps three through six until the requirements are met.

Images