High-pressure common rail system oil injection law self-adaptive observation method considering structural degradation

Abstract

The invention discloses a high-pressure common rail system oil injection rule self-adaptive observation method considering structure degradation, and belongs to the technical field of engine fuel system control, and the method considers the influence of oil injector structure degradation on model parameters on the basis of an oil injection process dynamic model, and adds degradation factors into the model. Then selecting a state variable, and constructing a state space model containing two outputs of rail pressure and a degradation factor; on the basis of a sequential extended Kalman filtering algorithm, a double-time-scale and double-output high-pressure common rail system oil injection estimator is designed, self-adaptive iterative correction is conducted on model parameters through measured rail pressure and recognized degradation factors, and real-time accurate estimation of the oil injection rule under the structure degradation condition is achieved. And meanwhile, the calculation real-time performance of an actual application process is also improved.

Description

Technical Field

[0001] The invention belongs to the technical field of engine fuel system control, and in particular relates to an online real-time observation method for the fuel injection amount of an engine high-pressure common rail fuel system. Background Art

[0002] As the central device of the engine, the fuel system's efficient and stable cyclic injection characteristics are crucial to ensuring the overall performance of the engine. The high-pressure common rail system provides strong support for the realization of green power with its high injection pressure and high response speed. However, it is currently difficult to obtain injection information in real time during actual operation. The injection amount of the existing common rail system mainly relies on the experimental calibration MAP map for open-loop control, resulting in limited injection amount control accuracy. Due to the complex hydraulic influences and changes in the working environment within the system, it is often difficult to ensure the consistency between actual and target injection. If the fuel injection information can be obtained in real time and closed-loop control of the injection amount can be achieved, the control accuracy and reliability of the injection process will be further improved.

[0003] Most existing studies obtain injection information by establishing a dynamic mathematical model and directly performing open-loop calculations. However, factors such as measurement noise, modeling errors, and improper initial condition settings may lead to deviations in the calculation results, or even difficulty in convergence. To overcome this limitation, the inventor has applied for a high-pressure common rail system injection prediction method based on a closed-loop observer. By constructing an injection state space model based on measurable rail pressure and using the error between the observed output and the measured output for closed-loop correction, the convergence of the observation results is ensured (Chinese patent: ZL202010577723.3, which has been authorized). On this basis, when the rail pressure changes over a large range, the system has a large nonlinearity. The inventor uses the extended Kalman filter algorithm to further optimize the observation process, update the feedback gain matrix in real time, and realize continuous iteration and rolling optimization of the injection state (Chinese patent: ZL202211287297.5, which has been authorized). The above methods are all for modeling and parameter identification of the initial state of the fuel system, but during the service life of the common rail system, its structural parameters will gradually degrade, especially the injector nozzle and needle valve assembly that are in a harsh environment of high temperature and high pressure for a long time, which is prone to carbon deposition in the spray hole, wear of the needle valve sealing surface and other faults, resulting in abnormal fuel injection. The above observation method proposed by the inventor has a certain robustness under the action of cyclic iterative correction, but when the injector structure is highly aged, ignoring the influence of structural degradation will still lead to deviations in the injection information estimation.

[0004] To this end, this patent proposes an adaptive observation method for injection of a high-pressure common rail system taking into account structural degradation, introduces a structural degradation factor into the injection model, and uses the rail pressure fluctuation characteristics to identify the degree of degradation, and establishes a state space model with dual outputs of rail pressure and degradation factor. Since the aging of the injector structure is a slow-changing process, the sampling frequencies of the two outputs of rail pressure and degradation factor are different. At the same time, in order to avoid the matrix inversion problem when solving the Kalman filter gain in the case of dual output, this patent is based on the sequential Kalman filter algorithm. Designed a dual-time scale, dual-output injection estimator, thereby realizing adaptive real-time estimation of the injection law. Summary of the invention

[0005] The purpose of the present invention is to provide an adaptive observation method for the injection law of a high-pressure common rail system considering structural degradation. Based on the dynamic model of the injection process, the method considers the influence of injector structural degradation on the model parameters and adds degradation factors to the model. Then, state variables are selected to construct a state space model containing two outputs, namely rail pressure and degradation factor. Since the two output sampling frequencies of the system are different and in order to improve the real-time performance of the calculation, the present invention designs a dual-time scale, dual-output high-pressure common rail system injection estimator based on the sequential extended Kalman filter algorithm.

[0006] The technical solution adopted by the present invention is specifically as follows:

[0007] An adaptive observation method for injection law of a high-pressure common rail system considering structural degradation, the specific steps of the method are:

[0008] Step 1: According to the fuel flow law and the influence mechanism of structural parameter degradation on the injection process, a dynamic model of the common rail system injection process considering structural degradation is established.

[0009] According to the common rail fuel flow process, the fuel continuity equation of the common rail is given:

[0010]

[0011] Where, p is the instantaneous pressure of the common rail pipe; Q pump The fuel flow rate supplied by the high-pressure fuel pump to the common rail pipe is negligible during the injection process; Q inj is the injection rate; Q loss is the fuel loss rate of the injector, that is, the oil return rate of the control chamber; E is the bulk elastic modulus of the fuel; V is the control volume of the common rail pipe.

[0012] By simplifying formula (1), the mathematical model between injection rate and instantaneous rail pressure can be obtained:

[0013]

[0014] In the formula, C lossis the fuel loss coefficient, which is the ratio of the return fuel volume to the injection fuel volume; C1 is the volume compensation coefficient. These two parameters can be identified using the injection fuel volume, return fuel volume and rail pressure data under normal conditions; V c is the common rail volume.

[0015] During long-term operation, the common rail system will wear and age, and its structural parameters will continue to degrade, which may lead to nozzle blockage or wear, needle valve jamming or sealing cone surface failure, thus causing abnormal fuel injection and affecting the fuel loss coefficient in formula (2). In order to describe the degree of abnormal injection caused by injector aging, the injector degradation factor β is introduced into formula (2), and the nonlinear dynamic model of the injection process considering structural degradation is obtained as follows:

[0016]

[0017] in

[0018]

[0019] In actual operation, the degradation factor β is identified by comparing the fluctuation characteristics of the rail pressure signal under normal and fault conditions.

[0020] Step 2: Establish an observable state space model and discretize it.

[0021] 1) To ensure the observability of the model, select the instantaneous rail pressure p and injection rate Q inj , the rate of change of injection rate and degradation factor β as state variables, namely

[0022]

[0023] Considering that the injection rate curve can be approximately regarded as a piecewise linear combination, that is, Approximate constant, and the degradation of injector structure is a slow process, β can also be approximate constant in a period of time, so there is Combined with formula (3), the nonlinear state space model of the fuel injection process is obtained as follows:

[0024]

[0025] Where y is the measured output of the system, namely the instantaneous rail pressure p and the degradation factor β.

[0026] 2) Discretize the continuous state space equation of equation (5) with a sampling step size of Δt. The state variable x(t) is k-1 The derivative at time can be approximately expressed as:

[0027]

[0028] tk ,t k-1 Substituting k and k-1, equation (6) can be written as:

[0029] x(k)=x(k-1)+Δt·f(x(k-1)) (7)

[0030] at this time,

[0031]

[0032] Combined with the output equation, the discrete state space model can be expressed as:

[0033]

[0034] in,

[0035] Step 3: Based on the sequential extended Kalman filter algorithm, an adaptive estimator of the injection law of the high-pressure common rail system with dual time scales and dual outputs is designed.

[0036] Considering the process noise w(k) and measurement noise v(k), equation (9) can be written as:

[0037]

[0038] w(k) and v(k) are independent zero-mean Gaussian white noises, and their covariance matrices are Q and R respectively. Where w(k) = [w1(k)w2(k)w3(k)w4(k)] T ; v(k) = [v1(k)v2(k)] T ; Q is a 4×4 dimensional diagonal matrix, Q=diag[q1 q2 q3q4], R is a 2×2 dimensional diagonal matrix, R=diag[r1 r2], and the corresponding elements in Q and R are taken according to the actual situation

[0039] According to the Kalman filter algorithm, the prior estimate of the state variable x(k) is The posterior estimate is First, linearize the nonlinear model (10). At time k, the posterior estimate of the state variable x(k) at time k-1 is In w(k-1)0=0, perform Taylor series expansion on x(k)=g(x(k-1),w(k-1)) in equation (11), omit the second-order and higher terms, and obtain the approximate linear expression of the state equation:

[0040]

[0041] Among them, A(k-1) and M(k-1) are x(k) in The Jacobian matrix at is calculated as follows:

[0042]

[0043] Substituting formula (8) into formula (12), we get:

[0044]

[0045] M(k-1) is a constant:

[0046]

[0047] Kalman filter optimal estimation includes two update processes: time update and measurement update. Initial value of error covariance P(0) + , and the noise covariance matrices Q and R, the calculation is performed according to the following time update and measurement update process, and the system state and covariance matrix are continuously updated in a cycle to finally obtain the optimal estimation result of the injection state. The specific steps are as follows:

[0048] 1) Time update phase:

[0049] First, the nonlinear model (9) is used to calculate the prior estimate

[0050]

[0051] Then, the Jacobian matrix A(k-1) is calculated according to formula (13) to obtain the prior estimated covariance matrix P(k) - :

[0052] P(k) - =A(k-1)P(k-1) + A(k-1) T +M(k-1)QM(k-1) T (16)

[0053] 2) Measurement update phase:

[0054] The two output measurements have different sampling rates, and dual time scales are used for updating in this stage. The fast-sampling rail pressure p is sampled in real time at Δt, and the slow-sampling degradation factor β is sampled at a longer sampling interval Δt. s =mΔt is identified by using the rail pressure signal characteristics (m is the ratio of fast sampling to slow sampling rate, which is an integer and can be set according to actual control requirements). Considering the influence of noise in the measured rail pressure, the degradation factor β is estimated a posteriori using filtering Extract rail pressure signal features.

[0055] A. When the sampling time k=n·mΔt (n is an integer greater than or equal to 1), the measurement data of both the rail pressure and the degradation factor dual outputs are available.

[0056] At this time, the dual-output system involves a matrix inversion process when updating the Kalman filter gain. In order to ensure the real-time calculation and avoid matrix inversion, the sequential Kalman filter algorithm is used to use all output measurement values ​​one by one to perform a posteriori estimation update. The components of the output equation in equation (10) are expressed as:

[0057]

[0058] Among them, y i (i=1,2) is the i-th output measurement value, y1(k)=p(k), y2(k)=β(k); C i (i=1,2) is the i-th row of the output matrix C, that is, C1=[1 0 0 0], C2=[0 0 0 1]; accordingly, let the Kalman filter gain be K(k)=[K1(k)K2(k)], where K i (k)(i=1,2) is the i-th column of the Kalman filter gain K(k).

[0059] First, select:

[0060]

[0061] Use the i-th output measurement value y at time k in sequence i (i=1,2) updates the posterior estimate, and the update equation is:

[0062]

[0063] B. When the sampling time k≠n·mΔt, the rail pressure measurement data is available.

[0064] At this time, the rail pressure measurement data y1(k) is used to measure and update the first three state variables, and the a posteriori estimate and error of the fourth state variable degradation factor remain unchanged. The output matrix of this stage becomes C f =[1 0 0].

[0065] definition The prior error corresponding to the first three state variables is P(k) - The upper left 3×3 matrix block is denoted as P f (k) - . Use P f (k) - and C f The Kalman filter gain, posterior estimate and posterior error covariance matrix are updated, and the equation is:

[0066]

[0067] P f (k) + =(IK(k)C f ) f (k) - (IK(k)C f ) T +r1K(k)K(k) T (25)

[0068] Among them, P f (k) + is the a posteriori error covariance corresponding to the first three state variables.

[0069] According to equations (21) and (24), we can get and In is the filtered rail pressure posterior estimation result, using the slow sampling interval Δt s The time domain and frequency domain features of the posterior estimated rail pressure are extracted to identify the output degradation factor β; the degradation factor β and the posterior estimated rail pressure are combined For the next iteration;

[0070] 3) The output system state posterior estimation results The current sampling time k can be obtained from is the optimal estimation result of the injection rate, In the injection phase (k∈[k0,k n ]) is integrated to obtain the optimal estimate of the injection amount in this injection phase

[0071]

[0072] Where k0 is the injection start time; k n The injection end time.

[0073] The advantages of the present invention are:

[0074] 1. This patent establishes a nonlinear model of fuel injection law based on instantaneous rail pressure according to the law of fuel flow, considers the influence of structural parameter degradation on the injection process, introduces degradation factors into the model, and uses rail pressure fluctuation characteristics to identify degradation factors. Based on this, an observable nonlinear state space model with rail pressure, injection rate, first-order derivative of injection rate and degradation factor as state variables is constructed, including rail pressure and degradation factor dual outputs.

[0075] 2. Taking model uncertainty and noise into consideration in the model, a real-time observation method for injection quantity based on extended Kalman filtering is proposed. In view of the different sampling frequencies of the two outputs of the system, a dual-time-scale Kalman filter is designed. The model parameters are adaptively iteratively corrected by measuring rail pressure and identifying degradation factors, so as to realize real-time and accurate estimation of injection law under structural degradation.

[0076] 3. The observation method proposed in this patent uses the rail pressure signal measured by the existing rail pressure sensor, without the need for additional measuring elements, and is easy to implement in engineering. At the same time, in order to avoid the matrix inversion process when solving the Kalman filter gain, when both output measurement values ​​are available at the slow sampling moment, the sequential Kalman filter method is used to update the posterior estimation, thereby improving the real-time calculation of the actual application process. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] In order to more clearly illustrate the technical solution of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, all other embodiments obtained without creative work are within the scope of protection of the present invention.

[0078] Figure 1 Schematic diagram of the adaptive observation method of the injection law of the high-pressure common rail system considering structural degradation;

[0079] Figure 2 Design process of adaptive Kalman filter for injection law;

[0080] Figure 3 Measure rail pressure and observe pressure curve;

[0081] Figure 4 Comparison of observed injection rate without correction and after adaptive correction;

[0082] Figure 5 Comparison of observed injection quantity without correction and after adaptive correction; DETAILED DESCRIPTION

[0083] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other implementations obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0084] Figure 1The present invention provides a schematic diagram of an adaptive observation method for the injection law of a high-pressure common rail system considering structural degradation. The method includes: collecting the pressure signal p of the common rail pipe of the high-pressure common rail system; inputting the rail pressure signal into the designed Kalman filter to obtain the posterior estimated rail pressure in real time. and injection rate Using slow sampling interval Δt s A posteriori estimate of rail pressure The time domain and frequency domain features are extracted, the output degradation factor β is identified, and input into the Kalman filter to realize the adaptive iterative correction of the degradation model parameters; in the injection stage, the injection rate is integrated to obtain the optimal estimation result of the injection amount.

[0085] Figure 2 The design process of the adaptive Kalman filter for the injection law of the high-pressure common rail system considering structural degradation is demonstrated. First, according to the fuel flow law and the influence of injector structural degradation, a nonlinear mathematical model of the injection law considering structural degradation is established; then, the state variables are selected to construct an observable state space model containing two outputs, rail pressure and degradation factor, and discretize them; on this basis, considering the different sampling frequencies of the dual outputs, a dual-time scale injection estimator is designed based on the extended Kalman filter algorithm, and when both output measurements are available, a posterior estimation update is performed in a sequential manner, finally achieving adaptive real-time and accurate estimation of the injection information.

[0086] The following are detailed descriptions:

[0087] Step 1: According to the fuel flow law and the influence mechanism of structural parameter degradation on the injection process, a dynamic model of the common rail system injection process considering structural degradation is established.

[0088] According to the common rail fuel flow process, the fuel continuity equation of the common rail is given:

[0089]

[0090] Where, p is the instantaneous pressure of the common rail pipe; Q pump The fuel flow rate supplied by the high-pressure fuel pump to the common rail pipe is negligible during the injection process; Q inj is the injection rate; Q loss is the fuel loss rate of the injector, that is, the oil return rate of the control chamber; E is the bulk elastic modulus of the fuel; V is the control volume of the common rail pipe.

[0091] By simplifying formula (1), the mathematical model between injection rate and instantaneous rail pressure can be obtained:

[0092]

[0093] In the formula, C lossis the fuel loss coefficient, which is the ratio of the return oil volume to the injection oil volume; C1 is the volume compensation coefficient. These two parameters can be identified using the injection oil volume, return oil volume and rail pressure data under normal conditions. This embodiment is for a four-stroke high-pressure common rail system, and the common rail pipe volume V is calculated based on the structural parameters. c =29061mm 3 , and using different working condition data, we can identify C1=5.048, C loss =0.2115.

[0094] During long-term operation, the common rail system will wear and age, and its structural parameters will continue to degrade, which may lead to nozzle blockage or wear, needle valve jamming or sealing cone surface failure, thus causing abnormal fuel injection and affecting the fuel loss coefficient in formula (2). In order to describe the degree of abnormal injection caused by injector aging, the injector degradation factor β is introduced into formula (2), and the nonlinear dynamic model of the injection process considering structural degradation is obtained as follows:

[0095]

[0096] in

[0097]

[0098] In actual operation, the degradation factor β is obtained by comparing the fluctuation characteristics of the rail pressure signal under normal and fault conditions.

[0099] Step 2: Establish an observable state space model and discretize it.

[0100] Step 2.1: To ensure the observability of the model, select the instantaneous rail pressure p and injection rate Q inj , the rate of change of injection rate and degradation factor β as state variables, namely

[0101]

[0102] Considering that the injection rate curve can be approximately regarded as a piecewise linear combination, that is, Approximate constant, and the degradation of injector structure is a slow process, β can also be approximate constant in a period of time, so there is Combined with formula (3), the nonlinear state space model of the fuel injection process is obtained as follows:

[0103]

[0104] Where y is the measured output of the system, namely the instantaneous rail pressure p and the degradation factor β.

[0105] Formula (5) is a nonlinear model. To verify its observability, linearization is achieved by deriving the Jacobian matrix, and then its observability is judged according to the rank of the observable matrix. The Jacobian matrix of formula (5) is obtained to obtain the system matrix A c for:

[0106]

[0107] Using the system matrix A c And the output matrix C c The observability matrix R is derived o , whose rank is 4, indicating that the nonlinear system is completely observable and a Kalman filter can be designed.

[0108] Step 2.2: Discretize the continuous state space equation of Equation (5) with a sampling step size of Δt. The state variable x(t) is k-1 The derivative at time can be approximately expressed as:

[0109]

[0110] t k ,t k-1 Substituting k and k-1, equation (7) can be written as:

[0111] x(k)=x(k-1)+Δt·f(x(k-1)) (8)

[0112] at this time,

[0113]

[0114] Combined with the output equation, the discrete state space model can be expressed as:

[0115]

[0116] in,

[0117] Step 3: Based on the sequential extended Kalman filter algorithm, a dual-output dual-time scale high-pressure common rail system injection law adaptive estimator is designed.

[0118] Considering the process noise w(k) and measurement noise v(k), equation (10) can be written as:

[0119]

[0120] w(k) and v(k) are independent zero-mean Gaussian white noises, and their covariance matrices are Q and R respectively. Where w(k) = [w1(k)w2(k)w3(k)w4(k)] T ; v(k) = [v1(k)v2(k)]T ; Q is a 4×4 dimensional diagonal matrix, Q=diag[q1 q2 q3q4], R is a 2×2 dimensional diagonal matrix, R=diag[r1 r2], and the corresponding elements in Q and R are taken according to the actual situation

[0121] According to the Kalman filter algorithm, the prior estimate of the state variable x(k) is The posterior estimate is First, linearize the nonlinear model (11). At time k, the posterior estimate of the state variable x(k) at time k-1 is In w(k-1)0=0, perform Taylor series expansion on x(k)=g(x(k-1),w(k-1)) in equation (11), omit the second-order and higher terms, and obtain the approximate linear expression of the state equation:

[0122]

[0123] Among them, A(k-1) and M(k-1) are x(k) in The Jacobian matrix at is calculated as follows:

[0124]

[0125] Substituting formula (9) into formula (13), we get:

[0126]

[0127] M(k-1) is a constant:

[0128]

[0129] Kalman filter optimal estimation includes two update processes: time update and measurement update. Initial value of error covariance P(0) + , and the noise covariance matrices Q and R, the calculation is performed according to the following time update and measurement update process, and the system state and covariance matrix are continuously updated in a cycle to finally obtain the optimal estimation result of the injection state. The specific steps are as follows:

[0130] 1) Time update phase:

[0131] First, the nonlinear model (10) is used to calculate the prior estimate

[0132]

[0133] Then, the Jacobian matrix A(k-1) is calculated according to formula (14) to obtain the prior estimated covariance matrix P(k) - :

[0134] P(k) - =A(k-1)P(k-1) + A(k-1) T +M(k-1)QM(k-1) T (17)

[0135] 2) Measurement update phase:

[0136] The two output measurements have different sampling rates, and dual time scales are used for updating in this stage. The fast-sampling rail pressure p is sampled in real time at Δt, and the slow-sampling degradation factor β is identified using the rail pressure signal characteristics at a longer sampling interval mΔt. In this embodiment, m is set to the ratio of the time interval between two injections to Δt. For a four-stroke four-injector high-pressure common rail system, at a speed of 1000rpm, the injection interval is 0.015s, Δt=0.1ms, then m=150. Taking into account the influence of noise in the measured rail pressure, the degradation factor β is estimated using a filtered a posteriori Extract rail pressure signal features.

[0137] A. When the sampling time k=n·mΔt (n is an integer greater than or equal to 1), the measurement data of both the rail pressure and the degradation factor dual outputs are available.

[0138] At this time, the dual-output system involves a matrix inversion process when updating the Kalman filter gain. In order to ensure the real-time calculation and avoid matrix inversion, the sequential Kalman filter algorithm is used to use all output measurement values ​​one by one to perform a posteriori estimation update. The components of the output equation in equation (11) are expressed as:

[0139]

[0140] Among them, y i (i=1,2) is the i-th output measurement value, y1(k)=p(k), y2(k)=β(k); C i (i=1,2) is the i-th row of the output matrix C, that is, C1=[1 0 0 0], C2=[0 0 0 1]; accordingly, let the Kalman filter gain be K(k)=[K1(k)K2(k)], where K i (k)(i=1,2) is the i-th column of the Kalman filter gain K(k).

[0141] First, select:

[0142]

[0143] Use the i-th output measurement value y at time k in sequence i (i=1,2) updates the posterior estimate, and the update equation is:

[0144]

[0145] B. When the sampling time k≠n·mΔt, the rail pressure measurement data is available.

[0146] At this time, the rail pressure measurement data y1(k) is used to measure and update the first three state variables, and the a posteriori estimate and error of the fourth state variable degradation factor remain unchanged. The output matrix of this stage becomes C f =[1 0 0].

[0147] definition The prior error corresponding to the first three state variables is P(k) - The upper left 3×3 matrix block is denoted as P f (k) - . Use P f (k) - and C f The Kalman filter gain, posterior estimate and posterior error covariance matrix are updated, and the equation is:

[0148]

[0149] P f (k) + =(IK(k)C f ) f (k) - (IK(k)C f ) T +r1K(k)K(k) T (26)

[0150] Among them, P f (k) + is the a posteriori error covariance corresponding to the first three state variables.

[0151] According to equations (22) and (25), we can get and In is the filtered rail pressure posterior estimation result, using the slow sampling interval Δt s The time domain and frequency domain features of the posterior estimated rail pressure are extracted to identify the output degradation factor β; the degradation factor β and the posterior estimated rail pressure are combined For the next iteration;

[0152] 3) The output system state posterior estimation results The current sampling time k can be obtained from is the optimal estimation result of the injection rate, In the injection phase (k∈[k0,k n ]) is integrated to obtain the optimal estimate of the injection amount in this injection phase

[0153]

[0154] In formula (27), k0 is the injection start time; k n The injection end time.

[0155] In order to verify the correction effect and observation accuracy of the adaptive observation method, a real-time online simulation study was conducted using the method proposed in the present invention under the conditions of rail pressure of 1200 bar and injection pulse width of 1.3 ms, taking the case of 30% nozzle blockage as an example. Initial value of error covariance P(0) + =I 4×4 , Q = diag[0.01 0.002 0.010.1] and R = diag[2.5 0.01]. No correction is performed before t = 0.7s, and adaptive correction starts after 0.7s.

[0156] Figure 3 To compare the measured rail pressure with the observed rail pressure, Figure 4 is the comparison curve of the observed injection rate without correction and after adaptive correction, Figure 5 The results of 10 consecutive injections before and after the adaptive correction are compared (the first five are uncorrected observations, and the last five are adaptively corrected observations). The results show that the uncorrected observed injection rate peak is higher than the actual injection rate, and the injection amount observation result is larger than the actual value, because the influence of the reduction of the injection flow area caused by the nozzle blockage is not considered. The average error of a single injection amount reaches 8.4%; the injection rate curve after adaptive correction is highly consistent with the actual injection rate, and the error of a single injection amount is only 3.1%, which is 5.3% higher than the observation accuracy before correction.

Claims

1. An adaptive observation method for injection law of a high-pressure common rail system considering structural degradation, characterized in that: The specific steps of this method are: Step 1: According to the fuel flow law and the influence mechanism of structural parameter degradation on the injection process, a dynamic model of the common rail system injection process considering structural degradation is established; The nonlinear dynamic model of the injection process considering structural degradation is: in p is the instantaneous pressure of the common rail pipe; β is the injector degradation factor; Q inj is the injection rate; V c is the volume of the common rail pipe; C loss is the fuel loss coefficient, which is the ratio of the return fuel volume to the injection fuel volume; C1 is the volume compensation coefficient; C loss The two undetermined parameters C1 and C2 are identified using the injection volume, return oil volume and rail pressure data under normal conditions; Step 2: Establish an observable state space model and discretize it; 1) To ensure the observability of the model, select the instantaneous rail pressure p and injection rate Q inj , the rate of change of injection rate and degradation factor β as state variables, namely According to the state variable and combined with equation (1), the nonlinear state space model of the fuel injection process is obtained as follows: Where y is the measured output of the system, i.e., the instantaneous rail pressure p and the degradation factor β; 2) Discretize the continuous state space equation of equation (3) with a sampling step size of Δt. The state variable x(t) is k-1 The derivative at time can be approximately expressed as: t k ,t k-1 Substituting k and k-1, formula (4) can be rewritten as: x(k)=x(k-1)+Δt·f(x(k-1)) (5) at this time, Combined with the output equation, the obtained discrete state space model is expressed as: in, Step 3: Based on the sequential extended Kalman filter algorithm, a dual-time-scale dual-output high-pressure common rail system injection law adaptive estimator is designed; Considering the process noise w(k) and measurement noise v(k), equation (7) can be rewritten as: w(k) and v(k) are independent zero-mean Gaussian white noises, and their covariance matrices are Q and R respectively; where w(k) = [w1(k)w2(k)w3(k)w4(k)] T ; v(k) = [v1(k)v2(k)] T ; Q is a 4×4 dimensional diagonal matrix, Q = diag[q1 q2 q3 q4], R is a 2×2 dimensional diagonal matrix, R=diag[r1 r2], and the corresponding elements in Q and R are taken according to the actual situation; According to the Kalman filter algorithm, the prior estimate of the state variable x(k) is The posterior estimate is First, the nonlinear model (8) is linearized; at time k, the posterior estimate of the state variable x(k) at time k-1 is In At w(k-1)0=0, perform Taylor series expansion on x(k)=g(x(k-1),w(k-1)) in equation (8), omit the second-order and higher terms, and obtain the approximate linear expression of the state equation: Among them, A(k-1) and M(k-1) are x(k) in The Jacobian matrix at is calculated as follows: Substituting formula (6) into formula (10), we get: M(k-1) is a constant: Kalman filter optimal estimation includes two update processes: time update and measurement update. Initial value of error covariance P(0) + , and the noise covariance matrices Q and R, the calculation is performed according to the following time update and measurement update process, and the system state and covariance matrix are continuously updated in a cycle to finally obtain the optimal estimation result of the injection state. The specific steps are as follows: 1) Time update phase: First, the nonlinear model (7) is used to calculate the prior estimate Then, the Jacobian matrix A(k-1) is calculated according to formula (11) to obtain the prior estimated covariance matrix P(k) - : 2) Measurement update phase: The two output measurements have different sampling rates, and dual time scales are used for updating in this stage; the fast-sampling rail pressure p is sampled in real time at Δt, and the slow-sampling degradation factor β is sampled at a longer sampling interval Δt s =mΔt, using rail pressure signal characteristics for identification; where m is the ratio of fast sampling to slow sampling rate, which is an integer and is set according to actual control requirements; A. When the sampling time k = n·mΔt (n∈N + ), the measurement data of both rail pressure and degradation factor outputs are available; At this time, the dual-output system involves a matrix inversion process when updating the Kalman filter gain. To ensure the real-time calculation and avoid matrix inversion, a sequential Kalman filter algorithm is used to use all output measurement values ​​one by one to perform a posteriori estimation update. The components of the output equation in equation (8) are expressed as: Among them, y i (i=1,2) is the i-th output measurement value, y1(k)=p(k), y2(k)=β(k); C i (i=1,2) is the i-th row of the output matrix C, that is, C1=[1 0 0 0], C2=[0 0 0 1]; Accordingly, let the Kalman filter gain be K(k)=[K1(k)K2(k)], where K i (k)(i=1,2) is the i-th column of Kalman filter gain K(k); First, select: Use the i-th output measurement value y at time k in sequence i (i=1,2) updates the posterior estimate, and the update equation is: B. When the sampling time k≠n·mΔt, the rail pressure measurement data is available; At this time, the rail pressure measurement data y1(k) is used to measure and update the first three state variables, and the a posteriori estimate and error result of the fourth state variable degradation factor remain unchanged from the prior value; definition The prior error corresponding to the first three state variables is P(k) - The upper left 3×3 matrix block is denoted as P f (k) - ; Use P f (k) - and C f The Kalman filter gain, posterior estimate and posterior error covariance matrix are updated, and the equation is: P f (k) + =(IK(k)C f )P f (k) - (IK(k)C f ) T +r1K(k)K(k) T (24) Among them, P f (k) + is the posterior error covariance corresponding to the first three state variables; According to equations (20) and (23), we can get and In is the filtered rail pressure posterior estimation result, using the slow sampling interval Δt s The time domain and frequency domain features of the posterior estimated rail pressure are extracted to identify the output degradation factor β; the degradation factor β and the posterior estimated rail pressure are combined For the next iteration; 3) The output system state posterior estimation results The current sampling time k can be obtained from is the optimal estimation result of the injection rate, In the injection phase (k∈[k0,k n ]) is integrated to obtain the optimal estimate of the injection amount in this injection phase In formula (25), k0 is the injection start time; k n The injection end time.

Images