Internal combustion engine cylinder pressure parameter correction method based on multi-loop iterative correction and internal combustion engine

Abstract

The application discloses a cylinder pressure parameter correction method based on multi-loop iterative correction and an internal combustion engine and belongs to the field of internal combustion engine combustion analysis and electronic control technology. According to the application, an observation model is established, a parameter estimation method is used to jointly or separately estimate sensor measurement offset estimation, polytropic index estimation and model constant estimation, and when there is an angle offset of the top dead center (TDC) of a cylinder, nonlinear quadratic estimation is carried out, and finally, an optimized cylinder pressure value is obtained. The application has strong adaptability to cylinder pressure changes caused by different engine operating states.

Description

Technical Field

[0001] This invention relates to the field of internal combustion engine combustion analysis and electronic control technology, and in particular to a method for correcting cylinder pressure parameters of an internal combustion engine based on multi-loop iterative correction, and an internal combustion engine. Background Technology

[0002] Cylinder pressure is one of the most crucial physical quantities characterizing the operation of an internal combustion engine. It is a key input for calculating the combustion heat release rate, indicating mean effective pressure (IMEP), evaluating combustion stability, and implementing closed-loop combustion control (such as CA50 control). The accuracy of cylinder pressure signal measurement directly affects the accuracy and reliability of all the aforementioned analysis and control functions. Currently, cylinder pressure measurement is mainly achieved through piezoelectric or piezoresistive sensors mounted on the cylinder head, and piston position (such as top dead center, TDC) is determined by crankshaft position sensors (such as encoders). However, in practical engineering applications, the accuracy of cylinder pressure signals has long been plagued by two key systematic errors: zero-point drift of the cylinder pressure sensor and identification offset of top dead center (TDC).

[0003] Under harsh working conditions of long-term high temperature, high pressure, and vibration, the zero-point output of the cylinder pressure sensor will slowly drift over time. This drift can cause a fixed deviation between the measured absolute pressure value and the actual value. This drift can be caused by various factors such as sensor material aging, temperature effects, and stress release during installation. On the other hand, factors such as installation errors of crankshaft position sensors (e.g., encoders), transmission system clearances, belt / chain tension variations, and even thermal deformation of the engine block can lead to an angular offset between the theoretically calculated top dead center position of the piston and the actual physical top dead center. These two errors are coupled and together distort the cylinder pressure-crankshaft rotation angle. p-α The compression stroke curve, in particular, severely affects the accuracy of the description of the compression stroke process. p-α The shape of the compression curve is crucial for calculating the compression ratio, evaluating cylinder sealing, and, most importantly, serving as a benchmark for combustion analysis (the calculation of combustion heat release rate heavily depends on the accuracy of the compression curve). Incorrect curve shape... and This directly leads to compression curve distortion, resulting in a series of problems such as "false heat release" in heat release rate calculations, misjudgment of combustion phase (e.g., CA50), and errors in thermal efficiency calculations, seriously affecting engine development, calibration, and fault diagnosis. For example, Middendorf et al. (Middendorf, H., et al. "A New Method for Cylinder Pressure Based Indicated Mean Effective Pressure (IMEP) Calculation with Consideration of Heat Transfer and TDC Drift.") SAE International Journal of Engines (2021) pointed out in their research that TDC offset has a significant impact on the calculation of key parameters such as IMEP and is an error source that cannot be ignored in high-precision analysis.

[0004] Traditionally, cylinder pressure sensor calibration typically relies on offline calibration, such as "atmospheric pressure calibration" when the engine is not running, or comparison with an external high-precision reference sensor under specific steady-state conditions. However, these methods cannot accommodate real-time drift generated by the sensor during operation. Correction for TDC offset is even more difficult, usually requiring complex specialized equipment (such as TDC sensors) or identification under specific conditions (such as reverse drag conditions or thermodynamic methods), making the process cumbersome and difficult to integrate into online control systems. Patent CN111999067A, "A method, device, and vehicle for calculating the physical dead center calibration of an engine," demonstrates a method for TDC calibration using the principle of compression / expansion line symmetry. Throughout the engine's lifespan, these offsets may change due to wear and aging, thus necessitating a method that can adaptively estimate and compensate for these errors online.

[0005] Existing research has attempted to use engine cycle data itself for online error estimation. A mainstream and physically sound approach is based on the polytropic process assumption of the compression stroke. During the pure compression phase from intake valve closing (IVC) to the start of combustion (SOC), ignoring complex heat transfer and leakage (or attributing their effects to an equivalent polytropic index), the in-cylinder process can be approximated as a polytropic process, satisfying the equations... p =C ,in For cylinder pressure, V For instantaneous cylinder volume, It is a variable index. C This is a constant determined by the initial state. When a sensor measures the offset... At that time, measure the pressure If the polytropic index is known... κ (For example, through experience or by calibrating to a fixed value such as 1.32), then the model is about and C It is linear and can be solved directly using the compressed segment data via the linear least squares (LS) method. and C This method is simple, fast, and computationally inexpensive, and was widely adopted in the early days. For example, the least squares-based cylinder pressure calibration method was used in patent KR20080018406A, "Cylinder Pressure Compensation Based on Least Squares Method". However, its drawbacks are quite obvious: the variable index... κ It is not constant; it varies with operating conditions such as engine speed, load, coolant temperature, EGR rate, and intake conditions, typically ranging from 1.25 (near isothermal, strong heat transfer) to 1.4 (near adiabatic, weak heat transfer). (Wittek et al., "A Comparative Study of Methods for Polytropic Exponent Estimation in Cylinder Pressure Analysis.") International Journal of Engine Research A comparative study (2023) confirms that fixed This will introduce model error, leading to Inaccurate estimations, especially under drastic changes in operating conditions, severely limit the accuracy of calibration. The preset accuracy.

[0006] To overcome fixed Limitations, further research will As an unknown parameter and Δp They are estimated together. At this point, the problem transforms into a nonlinear least squares problem. For example, Alberto di Gaeta, et al. (AKalman filter approach for the on-line estimation of the cylinder pressureoffset using a polytropic exponent model). Control Engineering Practice (2020) proposed using an extended Kalman filter (EKF) to... and Joint nonlinear estimation is performed as a state variable. However, such iterative optimization algorithms can be computationally burdensome for embedded systems, are sensitive to initial values, and are prone to getting trapped in local optima. Furthermore, most of these methods do not adequately consider the statistical characteristics of measurement noise and the coupling effects of TDC offset.

[0007] A more complex situation arises in actual compression processes, where heat transfer through the cylinder walls is not uniform and gas leakage may occur, leading to increased variability in the coefficient of performance. Its own properties may vary with crankshaft angle, especially in the early stages of compression (lower temperature, relatively strong heat transfer) and the later stages (higher temperature, relatively weaker heat transfer). Using a constant... Describing the entire process introduces systematic errors. (Liu et al., "Online Estimation of Cylinder Pressure Sensor Offset Using a Variable Polytropic Exponent Model Based on Recursive Least Squares.") IEEE Transactions on Industrial Electronics The study (2019) clearly points out that, considering The time-varying nature of the time-varying data is crucial for improving estimation accuracy, and a time-varying time-varying method based on recursive least squares is proposed. Model. Although some studies have proposed using more complex heat transfer models, these models have numerous parameters, making online identification difficult. Furthermore, a long-overlooked but significantly impactful factor is the zero-point drift of the cylinder pressure sensor. offset with TDC right p-α The influence of the curve is strongly coupled. A positive... This will shift the entire curve upwards, while a positive one... (Delayed TDC recognition) will cause the compressed line to twist and shift near the TDC. The combined effect of these two factors may differ from that of a single error source, making direct synchronous recognition less effective. and This can easily lead to unidentifiable parameters or increased estimation bias. Most existing methods address these two problems separately. Although Bares et al. and Chen et al. (Bares, P., et al. "A computationally efficient method for jointestimation of in-cylinder pressure offset and TDC position.") Mechanical Systems and Signal Processing(2022) Recent research has focused on developing computationally efficient joint estimation algorithms and robust joint estimation methods under disturbance conditions. However, significant challenges remain in balancing model complexity, time-varying parameters, coupling identification accuracy, and the need for online real-time computation. Recent explorations include those by Sánchez et al. (Sánchez, FJ, et al. "A Physics-Informed Neural Network Approach for In-Cylinder Pressure Reconstruction and Sensor Fault Diagnostics."). Engineering Applications of Artificial Intelligence The Physical Information Neural Network (PINN) method adopted in (2024) provides a new approach to dealing with such complex coupled and nonlinear problems.

[0008] In summary, the existing technology has the following main problems: (1) it relies on a fixed polynomial index. (1) The estimation accuracy is greatly affected by changes in working conditions; (2) The nonlinear parameter estimation algorithm is complex, and it is difficult to balance convergence and real-time performance; (3) The polytropic exponent is not fully considered. (4) Time-varying characteristics during compression; (5) Ineffective decoupling and joint estimation of cylinder pressure sensor zero-point drift. offset with TDC This leads to incomplete calibration. While current research has explored adaptive parameter estimation, joint estimation algorithms, and even data-driven methods that integrate with physical models, developing a method capable of adaptively estimating time-varying parameters remains a challenge. κ Robust and efficient joint estimation and It is applicable to cylinder pressure signal correction methods for online real-time operation and still has important theoretical value and engineering significance for improving the testing accuracy and advanced control capabilities of internal combustion engines. Summary of the Invention

[0009] The purpose of this invention is to address the technical deficiencies in the prior art by providing a method for correcting cylinder pressure parameters of internal combustion engines based on multi-loop iterative correction.

[0010] Another object of the present invention is to provide an internal combustion engine.

[0011] The technical solution adopted to achieve the purpose of this invention is: A method for correcting cylinder pressure parameters of an internal combustion engine based on multi-loop iterative correction includes the following steps: Step 1, Obtain initial cylinder pressure measurement value The corresponding crankshaft rotation angle Calculate the instantaneous cylinder volume corresponding to each crankshaft rotation angle.V ; Step 2: Establish an observation model for the cylinder pressure measurement and the instantaneous cylinder volume: ,in, The sensor measures the offset of the cylinder pressure, where κ is the polytropic index. C These are model constants; Step 3: Based on the observation model established in Step 2, use parameter estimation algorithms to estimate the sensor measurement offset jointly or separately. Estimates, variability index Estimation and model constants estimate; Step 4: During the estimation process, if there is an angular deviation at the top dead center (TDC) of the cylinder... At that time, the compression stroke range of the cylinder is divided into S consecutive sub-ranges, and angular offset is used. The instantaneous cylinder volume within each sub-interval is corrected, and an observation model for each sub-interval is established. An alternating iterative optimization strategy is adopted, and within each sub-interval, the values ​​obtained in step 3 are sequentially applied. Estimate and C estimate ， Optimize the polyvariate index, and then use the optimized polyvariate index to sequentially optimize the model constants. C and sensor measurement offset Finally, the angular offset was optimized using the optimized polytropic index, model constants, and sensor measurement offset. The process continues until the parameters converge, ultimately yielding the optimized sensor measurement offset estimate. and angle offset estimation ; Step 5, using the estimate from step 3 Estimate and / or estimate the sensor measurement offset in step 4 and angle offset estimation For the initial cylinder pressure measurement value and crankshaft angle The correction is performed to obtain the corrected cylinder pressure value. and the corrected crankshaft angle This is used for subsequent combustion analysis and control.

[0012] In the above technical solution, in step 1, the initial cylinder pressure measurement value and its corresponding crankshaft angle are obtained from the compression stroke range extracted from the engine's real-time data. The compression stroke is cut off from 30°CA after the intake valve closes to 20°CA before combustion begins, in order to avoid the effects of valve movement disturbance and initial combustion.

[0013] In the above technical solution, in step 2, ,in, This is the cylinder pressure measurement value at the moment the intake valve closes. This is the instantaneous cylinder volume at the moment the intake valve closes.

[0014] In the above technical solution, in step 3, when κ is known and is a constant, the parameter estimation algorithm is the linear least squares method, which is used to estimate the sensor measurement offset. And the model constant C, used for cylinder pressure correction, specifically: Initial cylinder pressure measurement value Discretized into n values, the observation model used in the linear least squares method is as follows:

[0015] Constructing a matrix ,in

[0016] in, To measure the initial cylinder pressure The first cylinder pressure measurement value after discretization. For the nth cylinder pressure measurement value, V 1 represents the instantaneous volume of the first cylinder after discretization. V n For the discretized first n Instantaneous volume of a cylinder For parameters, This is the regression matrix; The parameters The least squares estimate is: .

[0017] In the above technical solution, in step 3, when κ is unknown or changes, the parameter estimation algorithm is Newton's method (Newton iteration). Using Newton's method, the parameter is estimated by setting the value of κ. and C， The specific process is as follows: ： Step 3.1, take the initial cylinder pressure measurement value Discretize into n values ​​and define the objective function. for:

[0018] Among them, residual , For the discretized first i Initial cylinder pressure measurement value, For the first i The sensor measurement offset corresponding to each data point For the firsti The model constants corresponding to each data point For the first i The polyvariance index corresponding to each data point V i For the first i The instantaneous cylinder volume corresponding to each data point; Step 3.2: Set the initial conjectured value of the polyvariate index κ, and use the sensor measurement offset and model constants estimated by linear least squares method as initial values, and set the objective function... Solve using Newton's method iteratively: ( )

[0019] in, For the first i The objective function of the next iteration is... κ The first derivative of the gradient, For the first The objective function of the next iteration is... κ The second derivative of the gradient; Step 3.3, after each iteration, Apply boundary constraints until , To preset the tolerance, or to reach the maximum number of iterations, the optimal set of parameters is obtained: polytropic index. Estimate sensor measurement offset and model constants C。

[0020] In the above technical solutions, Second derivative .

[0021] In the above technical solution, step 4 specifically involves the following process: Step a, obtain the initial cylinder pressure measurement value in step 1. and the corresponding crankshaft rotation angle Discretize each value into n to obtain the cylinder pressure measurement value. and crankshaft angle ( i =0, 1... n The compression stroke of the cylinder is divided into S consecutive sub-intervals. When there is an angular offset at the top dead center of the cylinder... At that time, within each sub-interval, an alternating iterative optimization strategy is executed; Step b: Set an initial hypothetical value for the polytropic index, and use the sensor measurement offset estimated in step 3. and model constants COptimize the solution of each data point after discretization of the sub-interval. i The corresponding optimal polytropic index ; Step c, based on the optimal polytropic index in step b. And set the cylinder pressure corresponding to the reference point of the corresponding sub-interval. and instantaneous volume Calculate the discrete values ​​of each data point in each sub-interval. i Corresponding optimal model constants ; in, For the first j The first sub-interval i An optimized variability index for each data point.

[0022] Step d: Based on the optimal polyvariate index obtained in step b and the model constants calculated in step c, iterative optimization is performed using the following formula to obtain the discrete data points for each sub-interval. i Corresponding optimal sensor measurement offset : The observation model was revised as follows:

[0023] in, To measure the initial cylinder pressure Discretized i Individual cylinder pressure measurement value For the first j The number of sub-intervals, the first i The difference between the actual cylinder pressure and the virtual cylinder pressure at each data point. This is the corrected instantaneous cylinder volume; Step e, measure the offset based on the optimal sensor. Variable index and model constants Optimize the top dead center angle offset , define about Global fitting error function:

[0024] in, For iteration m The sensor measurement offset obtained after this step. In order to be in j The number of sub-intervals i Iterate through data pointsm The model constants obtained after this process In order to be in j The number of sub-intervals i Iterate through data points m The variable index obtained afterward.

[0025] Within a preset range, a one-dimensional search algorithm is used to find the... smallest The value is used to obtain the updated top dead center angle offset: ; Step f continues until the iteratively optimized sensor measurement offset, polytropic index, and angular offset are all less than the global convergence tolerance or the set maximum number of iterations is reached. The iteration cycle is then complete, and the sensor measurement offset estimate is obtained. and angle offset estimation .

[0026] In the above technical solution, in step 5, the corrected cylinder pressure value ,in, The result obtained in step 3 Estimate or obtained in step 4 The corrected crankshaft angle .

[0027] Another aspect of the present invention includes an electronic device, comprising a processor and a memory, wherein the memory is used to store program code for implementing the above method and to transmit the program code to the processor, wherein the processor is used to read cylinder pressure and crankshaft signals in real time according to the instructions in the program code, execute the method, and output corrected cylinder pressure data or directly use it for control decisions.

[0028] Another aspect of the invention includes a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the method, and the medium enables the algorithm to be easily deployed, ported, and updated.

[0029] Another aspect of the invention includes an internal combustion engine combustion analysis system, comprising: at least one cylinder pressure sensor for measuring cylinder pressure; at least one crankshaft position sensor for measuring crankshaft angular position; and the cylinder pressure signal processing device. This system can be configured as a standalone high-precision combustion analyzer, or integrated into an engine electronic control unit (ECU) as part of advanced control functions.

[0030] Another aspect of the invention includes an internal combustion engine comprising the aforementioned internal combustion engine combustion analysis system. This internal combustion engine possesses the capability for online self-calibration of its own cylinder pressure measurement system and advanced combustion analysis, laying the foundation for achieving more efficient and lower-emission intelligent combustion control.

[0031] Compared with the prior art, the beneficial effects of the present invention are: 1. Strong model adaptability and wide operating condition coverage: This invention provides a complete algorithm system ranging from simple linear models to complex piecewise nonlinear models, which can be flexibly selected according to accuracy requirements and computational resources. Algorithm 2 (unknown κ) and Algorithm 3 (piecewise κ) can adaptively estimate the polytropic exponent κ that varies with operating conditions and even with crankshaft angle, fundamentally overcoming the model error caused by a fixed κ. Testing shows that under various speeds from 600 to 1600 rpm and different load conditions, the algorithm can automatically identify reasonable variations of κ between 1.2 and 1.4, such as κ approaching 1.33 at low speeds and approaching 1.4 at high speeds, demonstrating strong adaptability to cylinder pressure changes caused by different engine operating states.

[0032] 2. Efficient and robust parameter estimation with controllable computational complexity: The proposed Newton method, combined with an alternating iterative optimization strategy, reduces nonlinear optimization to a one-dimensional search. It also utilizes analytical gradients and finite differences to approximate the second derivative, ensuring fast convergence while avoiding the computation of complex Hessian matrices. The algorithm has a simple structure, good numerical stability, and controllable computation time per iteration on typical embedded processors, making it highly suitable for online real-time operation. Tests show that the algorithm converges within 10 iterations under most operating conditions.

[0033] 3. Successfully decoupled and jointly estimated key offsets with high estimation accuracy: [This is related to...] and To address the problem of strong coupling, an innovative two-step identification method with alternating iterations was adopted, effectively breaking the mutual interference between parameters. In setting... For 2CA, In simulation experiments with a pressure of 1 bar, the joint estimation results of this method improved the RMSE of the cylinder pressure model from 0.1380 bar to 0.1017 bar, an improvement of 17.4%. Validation was performed on a dataset of 231 full-condition operating conditions covering 600-2150 rpm. The joint estimate RMSE is 0.1529CA, and the maximum error is only 0.2063CA, which confirms the high accuracy and strong robustness of the algorithm.

[0034] 4. Significantly improves the accuracy of cylinder pressure signals and combustion analysis, eliminating non-physical phenomena: through the analysis of Δp and Through precise compensation and accurate estimation of κ, a highly reliable cylinder pressure-rotation angle curve was obtained. Validation based on universal operating condition data shows that after joint compensation, the average RMSE of the virtual cylinder pressure model decreased from 0.946 bar to 0.876 bar, with an overall accuracy improvement of 7.35%. More importantly, the heat release rate calculated using this as input exhibits excellent physical consistency, effectively eliminating "spurious heat release" during the compression stroke.

[0035] 5. The algorithm boasts significant practical engineering value and is easy to integrate and promote: its modular design is clear, the physical meaning of the parameters is explicit, and their range is controllable (e.g., κ∈[1.2,1.4]). It provides a complete implementation process from data processing and volumetric calculation to core estimation. The algorithm has been successfully validated through MATLAB / Simulink simulation and offline batch data processing (over 231 operating points). The computational load has been optimized, and it has the potential to be directly ported to automotive ECUs or dedicated combustion analysis equipment for online real-time operation, demonstrating broad application prospects. Attached Figure Description

[0036] Figure 1 This is a flowchart illustrating the overall process of the method described in this invention.

[0037] Figure 2 The figure shows the fitting effect between the cylinder pressure obtained in Example 3 and the actual cylinder pressure, where the dashed line represents the virtual cylinder pressure and the solid line represents the actual cylinder pressure.

[0038] Figure 3 The cylinder pressure-rotation angle curves are shown for Example 3 before and after TDC offset compensation and the actual cylinder pressure under the conditions of 1300 rpm and 140 mg / hub fuel injection.

[0039] Figure 4 The calibration error diagrams for the sensor measurement offset Δp obtained in Examples 1-3 at different engine speeds are shown, where Alg 1 Error is for Example 1, Alg 2 Error is for Example 2, Alg 3 Error is for Example 3, and Zero Error is the 0 error line.

[0040] Figure 5 The graph shows a comparison between the heat release rate curve obtained using the traditional heat release rate formula and the heat release rate curve of Example 3, where SOI is the injection point and SOC is the combustion initiation point. Detailed Implementation

[0041] The present invention will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0042] Example 1 (Real-time estimation algorithm for engine cylinder pressure sensor measurement offset, Algorithm 1) This embodiment is designed for applications with extremely high real-time requirements and where a "working condition-variable index κ" mapping table has been established through extensive calibration (e.g., online misfire detection or roughness calculation modules for certain mass-produced engines). Algorithm 1 is used to achieve rapid online estimation of the cylinder pressure sensor measurement offset Δp.

[0043] A method for correcting cylinder pressure parameters of an internal combustion engine includes the following steps: Step 1: During engine operation, the engine electronic control unit (ECU) collects the cylinder pressure signal and its corresponding crankshaft signal in real time, and extracts the measurement data during the compression stroke from intake valve closing to combustion start to obtain the initial cylinder pressure measurement value. The corresponding crankshaft rotation angle Discretize it to obtain A discrete point sequence ,in Based on the engine's geometric parameters (such as connecting rod ratio, bore, stroke, compression ratio, etc.), calculate each crankshaft rotation angle. The corresponding instantaneous volume of the cylinder ; Step 2: Establish an observation model for the cylinder pressure measurement and the instantaneous cylinder volume: Among them, the cylinder pressure sensor measures the offset. κ is the polytropic index, and C is the model constant determined by the initial state; Step 3: Based on the observation model established in Step 2, calculate the current real-time engine speed. ,load and coolant temperature By querying a pre-calibrated two-dimensional or three-dimensional mapping table for key operating parameters, the recommended polyvariate index κ corresponding to the current operating condition is obtained, thus yielding the volumetric term sequence. At this point, the observation model is:

[0044] Step 4, construct the linear model matrix ,in:

[0045] in, To measure the initial cylinder pressure The first cylinder pressure measurement value after discretization. For the nth cylinder pressure measurement value, V 1 represents the instantaneous volume of the first cylinder after discretization. V n For the discretized first n Instantaneous volume of a cylinder For parameters, This is the regression matrix; To measure the offset of the sensor, These are constants related to the polytropic compression process. Let be the parameters, and according to the least squares principle, make the sum of squared residuals... The minimum parameter estimate is:

[0046] Step 5: In the ECU embedded system, the solution can be obtained directly using analytical methods or by employing efficient numerical methods (such as Cholesky decomposition). The first element of the solution vector is taken as the initial estimate of the sensor measurement offset, denoted as... , ;pass Calculations yielded Preferably, the sensor offset can be smoothed by filtering: to suppress fluctuations between cycles, continuous... A cycle (e.g.) The estimated initial sensor measurement offset is then averaged or subjected to a first-order low-pass filter to obtain the final smoothed sensor measurement offset used for real-time correction. Preferably, the cutoff frequency of the low-pass filter can be adjusted according to the engine's main excitation frequency (such as half of the ignition frequency).

[0047] Step 6, Real-time correction of cylinder pressure signal: using the smoothed offset The original cylinder pressure measurement signal is compensated online to obtain the corrected cylinder pressure value. : .

[0048] The estimation method in this embodiment has low computational complexity and low latency, and can effectively track the slow offset of the sensor, making it suitable for control loops that require real-time pressure feedback.

[0049] Example 2 (Estimating cylinder pressure sensor measurement offset and in-cylinder polytropic index using Newton's method, Algorithm 2) This embodiment is used for high-precision combustion analysis during engine bench testing and the R&D phase, or for online advanced control with higher calibration accuracy requirements (such as closed-loop control based on cylinder pressure), to estimate sensor measurement offset Δp and model constants. C And the variability index κ.

[0050] Step 1 is the same as step 1 in Example 1, resulting in... A discrete point sequence To reduce the impact of cyclic fluctuations, data averaged over multiple consecutive cycles is used to calculate each crankshaft angle based on engine geometry parameters (such as connecting rod ratio, bore, stroke, compression ratio, etc.). The corresponding instantaneous volume of the cylinder ; Step 2: Establish an observation model for the cylinder pressure measurement and the instantaneous cylinder volume: ,in, The sensor measures the offset for cylinder pressure. κ It is a variable index. C These are model constants; Step 3 specifically includes the following steps: Step 3.1, Initial guess of the polytropic index: Preset tolerance Maximum number of iterations: ; Polytropic exponential physical boundary: , For the variable index of the current iteration , i =0, 1…20, calculate the sequence of volume power terms. The initial sensor measurement offset estimated using the linear least squares method in step 3 of Example 1 is denoted as... The model constant is denoted as .

[0051] Step 3.2, Calculate the residuals and objective function values: Define the objective function (sum of squared residuals): Simplified to one-dimensional function : Calculate the residual of the current iteration and

[0052] objective function The value is used to measure the current parameter (polyvariance index). The cylinder pressure sensor measures the offset. and model constants The goodness of fit of the data.

[0053] Step 3.3, Calculate the objective function Regarding the variable index The gradient (first derivative) :

[0054] Step 3.4, calculate the objective function pair Second derivative approximation: The Hessian matrix (second derivative) is approximated using Newton's method. If it is the first iteration ( If the finite difference cannot be calculated, then a method with a fixed learning rate can be used. The gradient descent method performs the first step of the update: Then, skip directly to step 3.6 to process boundary constraints and start the next loop.

[0055] like Then the second derivative is approximated using the first-order difference quotient:

[0056] Step 3.5, Update the volatility index The update strategy is selected based on the value of the second derivative. A small positive number is set. (For example This is to prevent numerical overflow.

[0057] like Update using Newton's method:

[0058] Otherwise, when the second derivative approaches zero, revert to gradient descent for updating:

[0059] Step 3.6, Apply physical boundary constraints: To ensure the physical reasonableness of the estimation results, apply physical boundary constraints to the updated... Apply constraints.

[0060] Step 3.7, check convergence conditions: determine whether the iteration meets the stopping conditions.

[0061] If the variability index passes i After the second iteration, the change is less than the tolerance. : If the convergence is achieved, then the optimized result is considered to be... , , And then break out of the iteration loop.

[0062] If the number of iterations achieve If the loop still fails to converge, the loop is forcibly terminated, and the polytropic index obtained from the last iteration is output as the result.

[0063] Step 4, Output the jointly estimated and optimized parameters: After the iteration loop ends, output the final set of optimal parameters: sensor measurement offset. Variable index and model constants C .

[0064] Step 5, Correct cylinder pressure based on joint estimation results: Measure the offset using the optimal sensor estimated in Step 4. Online calibration of the cylinder pressure measurement signal yields a more physically accurate cylinder pressure value. :

[0065] This example demonstrates the core steps of Newton's method. In actual programming, special cases in the early stages of iteration and numerical stability issues need to be addressed. The algorithm typically converges within 5-10 iterations.

[0066] Example 3: Comprehensive calibration based on joint estimation of Algorithm 3 and TDC This embodiment is used for the highest precision combustion analysis, such as performance evaluation in scientific research or extreme operating conditions, while estimating the segmented polytropic index. and angular offset .

[0067] Step 1: The piecewise polytropic index and TDC offset joint calibration algorithm module estimates the cylinder pressure sensor measurement offset by modeling the piecewise compression process and using the golden section search method. Multivariable index of each compression section and top dead center angle offset To achieve high-precision comprehensive correction of cylinder pressure signals, the following steps are included: Step 1.1, Data Preparation and Segmentation Parameter Setting: Based on Step 1 of Example 1, the following steps are obtained... The compression stroke range of the cylinder is divided into S consecutive sub-ranges, preferably... And define the crankshaft angle position of the sub-interval boundary relative to the top dead center of combustion: Set the upper endpoint offset search range: .

[0068] Step 1.2, Parameter Initialization: Set the initial values ​​for all parameters to be estimated. The initialization parameters include: Top dead center angle offset: Sensor measures offset: Variation index for each sub-interval: ,in, Step 1.3, execute the outer-layer alternating iterative optimization strategy: set the global convergence tolerance. Let the number of iterations be... m ,from Begin by performing the following alternating optimization steps: use the top dead center angle offset of the current iteration. Correct the crankshaft angle for each data point The instantaneous cylinder volume is corrected to:

[0069] in, This is a function for calculating the instantaneous cylinder volume based on engine geometry parameters.

[0070] The specific process of the alternating iterative optimization strategy is as follows: (The sensor measurement offset obtained in Example 1 is denoted as,) And model constants, parallel optimization of polytropic exponents in three sub-intervals ): Step a, define the sub-interval index function: for each data point Based on its corrected crankshaft angle Determine the sub-interval , where: subinterval 1: Subinterval 2: Subinterval 3: .

[0071] Step b, perform sub-interval optimization in parallel: for each sub-interval Extract the set of all data points belonging to this sub-interval. .by The initial polytropic index guess value is denoted as the sensor measurement offset obtained in Example 1. Given the model constants, Newton's method (the algorithm in Example 2) is applied to the data of this sub-interval to optimize the solution of the optimal polytropic index for this sub-interval. .

[0072] Step c, update the sub-interval parameters: then based on the polytropic index of each sub-interval optimized in step b. Cylinder pressure at a pre-selected reference point Instantaneous cylinder volume at reference point (For example, the midpoint of this sub-interval can be used as a reference point, and its corresponding cylinder pressure) With instantaneous volume ), calculate the model constants for each subinterval. :

[0073] Step d: Fix the segmented model and update the sensor measurement offset: Use step b to obtain the polytropic index of the sub-interval. and model constants The discrete data points for each sub-interval are obtained by iterative optimization using the following formula. i Corresponding optimal sensor measurement offset : The observation model was revised as follows:

[0074] in, To measure the initial cylinder pressure Discretized i Individual cylinder pressure measurement value For the first j The number of sub-intervals, the first i The difference between the actual cylinder pressure and the virtual cylinder pressure at each data point. This is the corrected instantaneous cylinder volume; Step e, fix the cylinder pressure parameters and optimize the top dead center angle offset: measure the offset based on the optimal sensor. Variable index and model constants Optimize the top dead center angle offset , define about Global fitting error function:

[0075] In the preset Above, use the golden section method (or other one-dimensional search algorithm) to find the... smallest The golden section method continuously compares the function values ​​of two interior points within a sub-interval, narrowing the interval by a fixed ratio (approximately 0.618) until the interval length is less than a preset tolerance; thus obtaining the optimized angle offset. Step f, check global convergence: calculate the changes in each parameter after this iteration:

[0076] If the changes in all parameters are less than the global convergence tolerance, i.e. If the algorithm converges, then we exit the outer loop. Otherwise, let... Return to step c and continue iterating.

[0077] Step 4, Output the joint calibration optimal parameters: After the outer iterative loop converges or reaches the maximum number of iterations, output the final estimated set of optimal parameters: Optimal sensor measurement offset: ; Optimal variability index for each sub-interval: ,in ; Optimal top dead center angle offset: .

[0078] Step 1.5, Perform comprehensive cylinder pressure signal correction: Using the optimal parameters obtained in Step 1.4, perform joint correction on the original cylinder pressure measurement signal and crankshaft angle signal:

[0079] Corrected signal and It eliminates sensor measurement offset and top dead center phase error, providing a more accurate data basis for engine combustion analysis.

[0080] This example demonstrates the most complex joint estimation process. In practice, it may not be necessary to run the complete piecewise model every time; instead, algorithm combinations can be flexibly selected based on data quality and accuracy requirements.

[0081] At 1300 rpm and an injection rate of 140 mg / hub, the virtual cylinder pressure corrected in Example 3 was... The results were obtained by fitting the actual cylinder pressure. Figure 2 As shown, the virtual cylinder pressure can closely match the actual cylinder pressure.

[0082] like Figure 3 As shown, the cylinder pressure after TDC offset compensation in Example 3 is closer to the actual cylinder pressure.

[0083] The calibration errors of the sensor measurement offset Δp obtained in Examples 1-3 at different engine speeds are as follows: Figure 4 As shown, by Figure 4 It can be seen that the sensor measurement offset Δp estimation obtained in Example 3 has the best effect. For example... Figure 4 As shown, Example 3 has the highest estimation accuracy for Δp ​​under all test conditions, with a root mean square error (RMSE) of 0.1831 bar, which is significantly better than the algorithm of Example 1 (0.5599 bar).

[0084] like Figure 5 As shown, when the engine speed is 1850 rpm, the exothermic index is set to 1.35, and the exothermic curve is simulated using the heat release rate formula. The algorithm of Example 3 is used to simulate the heat release curve, demonstrating that the algorithm of Example 3 can effectively solve the problem that the heat release rate calculated using the traditional heat release rate formula is not zero in the compression stage. Example 4 An engine cylinder pressure signal processing device, in terms of hardware, can be a separate circuit board integrating a high-speed ADC (for acquiring analog cylinder pressure signals), a digital input interface (for receiving digital signals from the crankshaft encoder), a high-performance microprocessor (such as TI's C2000 series DSP or ARM Cortex-R series), and memory. In terms of software, program code containing the algorithms described in this invention (e.g., Algorithm 1 or Algorithm 2) is burned into the microprocessor's Flash memory. The device receives sensor signals in real time, executes estimation and correction algorithms, and outputs the corrected digital cylinder pressure signal and estimated parameters (such as Δp, κ) via CAN bus or Ethernet for display by a host computer or use by the ECU.

[0085] Example 5 A computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement a method, the medium enabling the algorithm to be easily deployed, ported, and updated.

[0086] Example 6 An internal combustion engine combustion analysis system includes: at least one cylinder pressure sensor for measuring cylinder pressure; at least one crankshaft position sensor for measuring crankshaft angular position; and the cylinder pressure signal processing device. This system can function independently as a high-precision combustion analyzer, or it can be integrated into the engine electronic control unit (ECU) as part of advanced control functions.

[0087] Example 7 An internal combustion engine includes the aforementioned internal combustion engine combustion analysis system. This internal combustion engine possesses the capability for online self-calibration of its own cylinder pressure measurement system and advanced combustion analysis, laying the foundation for achieving intelligent combustion control with higher efficiency and lower emissions.

[0088] The above description is only a preferred embodiment of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for correcting cylinder pressure parameters of an internal combustion engine based on multi-loop iterative correction, characterized in that, Includes the following steps: Step 1, Obtain initial cylinder pressure measurement value The corresponding crankshaft rotation angle Calculate the instantaneous cylinder volume corresponding to each crankshaft rotation angle. V ; Step 2: Establish an observation model for the cylinder pressure measurement and the instantaneous cylinder volume: ,in, The sensor measures the offset of the cylinder pressure, where κ is the polytropic index. C These are model constants; Step 3: Based on the observation model established in Step 2, use parameter estimation algorithms to estimate the sensor measurement offset jointly or separately. Estimates, variability index Estimation and model constants estimate; Step 4: During the estimation process, if there is an angular deviation at the top dead center of the cylinder... At that time, the compression stroke range of the cylinder is divided into S consecutive sub-ranges, and angular offset is used. The instantaneous cylinder volume within each sub-interval is corrected, and an observation model for each sub-interval is established. An alternating iterative optimization strategy is adopted, and within each sub-interval, the values ​​obtained in step 3 are sequentially applied. Estimate and C estimate ， Optimize the polyvariate index, and then use the optimized polyvariate index to sequentially optimize the model constants. and sensor measurement offset Finally, the angular offset was optimized using the optimized polytropic index, model constants, and sensor measurement offset. The process continues until the parameters converge, ultimately yielding the optimized sensor measurement offset estimate. and angle offset estimation ; Step 5, using the estimate from step 3 Estimate and / or estimate the sensor measurement offset in step 4 and angle offset estimation For the initial cylinder pressure measurement value and crankshaft rotation angle The correction is performed to obtain the corrected cylinder pressure value. and the corrected crankshaft angle This is used for subsequent combustion analysis and control.

2. The method for correcting cylinder pressure parameters of an internal combustion engine according to claim 1, characterized in that, In step 2, ,in, This is the cylinder pressure measurement value at the moment the intake valve closes. This is the instantaneous cylinder volume at the moment the intake valve closes.

3. The method for correcting cylinder pressure parameters of an internal combustion engine according to claim 1, characterized in that, In step 3, when κ is known and is a constant, the parameter estimation algorithm is the linear least squares method, which is used to estimate the sensor measurement offset. and model constants C Used for cylinder pressure calibration; preferably, the initial cylinder pressure measurement value is used. Discretized into n values, the observation model used by the linear least squares method is as follows: Constructing a matrix ,in in, To measure the initial cylinder pressure The first cylinder pressure measurement value after discretization. For the nth cylinder pressure measurement value, V 1 represents the instantaneous volume of the first cylinder after discretization. V n For the discretized first n Instantaneous volume of a cylinder For parameters, This is the regression matrix; Preferably, the parameters The least squares estimate is: 。 4. The method for correcting cylinder pressure parameters of an internal combustion engine according to claim 3, characterized in that, In step 3, when κ is unknown or changes, the parameter estimation algorithm is Newton's method. Using Newton's method, the parameter is estimated by setting the value of κ. and C The preferred method, specifically, is as follows: Step 3.1, take the initial cylinder pressure measurement value Discretize into n values ​​and define the objective function. for: Among them, residual , For the discretized first i Initial cylinder pressure measurement value, For the first i The sensor measurement offset corresponding to each data point For the first i The model constants corresponding to each data point For the first i The polyvariance index corresponding to each data point V i For the first i The instantaneous cylinder volume corresponding to each data point; Step 3.2: Set the initial conjectured value of the polyvariate index κ, and use the sensor measurement offset and model constants estimated by linear least squares method as initial values, and set the objective function... Solve using Newton's method iteratively: （ ） in, For the first i The objective function of the next iteration is... κ The first derivative of the gradient, For the first The objective function of the next iteration is... κ The second derivative of the gradient, For the first i The variable index of the next iteration. For the first i +1 iteration of the polytropic index; Step 3.3, after each iteration, Apply boundary constraints until , To preset the tolerance, or to reach the maximum number of iterations, the optimal set of parameters is obtained: polytropic index. Sensor measures offset and model constants C ; Preferred, Second derivative .

5. The method for correcting cylinder pressure parameters of an internal combustion engine according to claim 1, characterized in that, In step 4, the specific process is as follows: Step a, obtain the initial cylinder pressure measurement value in step 1. and the corresponding crankshaft rotation angle Discretize each value into n to obtain the cylinder pressure measurement value. and crankshaft angle ( i =0, 1... n The compression stroke of the cylinder is divided into S consecutive sub-intervals. When there is an angular offset at the top dead center of the cylinder... At that time, within each sub-interval, an alternating iterative optimization strategy is executed; Step b: Set an initial hypothetical value for the polytropic index, and use the sensor measurement offset estimated in step 3. and model constants C Optimize the solution of each data point after discretization of each sub-interval i The corresponding optimal polytropic index ; Step c, based on the optimal polytropic index in step b. And set the cylinder pressure corresponding to the reference point of the corresponding sub-interval. and instantaneous volume Calculate the discrete values ​​of each data point in each sub-interval. i Corresponding optimal model constants ; in, For the first j The first sub-interval i An optimized variability index for each data point; Step d, based on the optimal polytropic index in step b. The optimal model constants obtained in step c The discrete data points for each sub-interval are obtained through iterative optimization using the following formula. i Corresponding optimal sensor measurement offset : The observation model was revised as follows: in, To measure the initial cylinder pressure Discretized i Individual cylinder pressure measurement value For the first j The first sub-interval i The difference between the actual cylinder pressure and the virtual cylinder pressure at each data point. This is the corrected instantaneous cylinder volume; Step e, measure the offset based on the optimal sensor. Variable index and model constants Optimize angle offset , define about Global fitting error function: in, For iteration m The sensor measurement offset obtained after this step. In the first j The number of sub-intervals i Iterate through data points m The model constants obtained after this process In the first j The number of sub-intervals i Iterate through data points m The subsequent variable index; Within a preset range, a one-dimensional search algorithm is used to find the... smallest The value yields the optimized angle offset: ； Step f continues until the iteratively optimized sensor measurement offset, polytropic index, and angular offset are all less than the global convergence tolerance or the set maximum number of iterations is reached. The iteration cycle is then complete, and the sensor measurement offset estimate is obtained. and angle offset estimation .

6. The method for correcting cylinder pressure parameters of an internal combustion engine according to claim 1, characterized in that, In step 5, the corrected cylinder pressure value ,in, The result obtained in step 3 Estimate or obtained in step 4 The corrected crankshaft angle 。 7. An electronic device, characterized in that, The system includes a processor and a memory. The memory is used to store program code for implementing the internal combustion engine cylinder pressure parameter correction method as described in any one of claims 1 to 6, and to transmit the program code to the processor. The processor is used to read cylinder pressure and crankshaft signals in real time according to the instructions in the program code, execute the internal combustion engine cylinder pressure parameter correction method, and output the corrected cylinder pressure data or directly use it for control decisions.

8. A computer-readable storage medium, characterized in that, The medium stores computer-executable instructions, which, when executed by a processor, are used to implement the internal combustion engine cylinder pressure parameter correction method as described in any one of claims 1 to 6.

9. An internal combustion engine combustion analysis system, characterized in that, It includes at least one cylinder pressure sensor for measuring cylinder pressure; at least one crankshaft position sensor for measuring crankshaft angular position; and the electronic device as described in claim 7.

10. An internal combustion engine, characterized in that, The internal combustion engine combustion analysis system as described in claim 9 is included.

Images