Control device and control method

Abstract

To reduce variation in an actual injection amount of fuel due to deterioration with time.SOLUTION: A control device (10) includes: a prediction section (22) for outputting a predicted injection amount (Yp) when a common rail system (50) is new by predicting, on the basis of current operation conditions (x1-xn) of the common rail system (50), a target injection amount calculated by isochronous control when the common rail system (50) is new; a storage section (232) for storing a correction map (23T) associated with a correction amount (Qc) with respect to the predicted injection amount (Yp); a correction amount calculating section (231) for calculating the correction amount (Qc) in correspondence with the predicted injection amount (Yp) on the basis of the correction map (23T); and a correction value calculating section (25) for outputting a corrected target injection amount (Ya) by adding the correction amount (Qc) to a current target injection amount (Y) calculated by isochronous control.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a control device and a control method. [Background technology]

[0002] Common rail systems that supply fuel to diesel engines are sometimes controlled by isochronous control (see, for example, Patent Document 1). With isochronous control, the target engine speed can be maintained by adjusting the target fuel injection amount even if the load increases or decreases. Therefore, isochronous control is often used in industrial (OHW: Off-Highway) vehicles such as excavators and wheel loaders, which are prone to load increases or decreases. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2007-198299 Summary of the Invention [Problem to be solved by the invention]

[0004] In a common rail system, the actual amount of fuel injected can vary due to injector deterioration over time, even if the injector's energization time and common rail pressure remain the same. While the actual injection amount may increase, it usually decreases. When the actual injection amount decreases, the target injection amount calculated by isochronous control increases compared to when the engine was new, in order to maintain a certain load at a certain engine speed.

[0005] To prevent excessive torque increases, an upper limit is usually set for the target injection quantity. When deterioration progresses and the target injection quantity reaches this limit, it becomes impossible to increase the actual injection quantity any further, limiting the high-load operations that were possible when the engine was new. In addition, the rail pressure of the common rail is controlled based on a rail pressure map that determines the target rail pressure of the common rail in relation to the target injection quantity. If the target injection quantity increases from when the engine was new, the reading position of the rail pressure map changes, which changes the target rail pressure and affects emissions.

[0006] The present invention aims to reduce fluctuations in the actual fuel injection amount due to deterioration over time. [Means for solving the problem]

[0007] One aspect of the present invention is a control device (10) that adjusts a fuel injection amount in a common rail system (50) by isochronous control. The control device (10) includes: a prediction unit (22) that predicts a target injection amount calculated by the isochronous control when the common rail system (50) is new based on current operating conditions (x1 to xn) of the common rail system (50) and outputs a predicted injection amount (Yp) when the common rail system (50) is new; a storage unit (232) that stores a correction map (23T) in which a correction amount (Qc) is associated with the predicted injection amount (Yp); a correction amount calculation unit (231) that calculates the correction amount (Qc) corresponding to the predicted injection amount (Yp) based on the correction map (23T); and a correction value calculation unit (25) that outputs a corrected target injection amount (Ya) by adding the correction amount (Qc) to the current target injection amount (Y) calculated by the isochronous control.

[0008] Another aspect of the present invention is a control method for adjusting a fuel injection amount in a common rail system (50) by isochronous control. The control method includes the steps of: predicting a target injection amount calculated by the isochronous control when the common rail system (50) is new, based on current operating conditions (x1 to xn) of the common rail system (50), and outputting a predicted injection amount (Yp); calculating a correction amount (Qc) corresponding to the predicted injection amount (Yp), based on a correction map (23T) in which the correction amount (Qc) is associated with the predicted injection amount (Yp); and outputting a corrected target injection amount (Ya) by adding the correction amount (Qc) to the current target injection amount (Y) calculated by the isochronous control. [Effects of the Invention]

[0009] According to the present invention, it is possible to reduce fluctuations in the actual fuel injection amount due to deterioration over time. [Brief explanation of the drawings]

[0010] [Figure 1] 1 is a diagram illustrating a control device and a common rail system according to an embodiment of the present invention; [Figure 2] FIG. 2 is a block diagram showing the configuration of a correction unit. [Figure 3] 4 is a flowchart showing a process performed by the control device to control the amount of fuel injection. [Figure 4A] FIG. 10 is a diagram illustrating an example of a correction map. [Figure 4B] FIG. 10 is a diagram illustrating an example of an updated correction map. [Figure 5] 10 is a graph showing the correlation between the current target injection amount, the predicted injection amount for a new product, and the corrected target injection amount. DETAILED DESCRIPTION OF THE INVENTION

[0011] An electronic control unit (ECU) that controls fuel supply to a diesel engine mounted on an OHW vehicle will be described below as one embodiment of a control device of the present invention with reference to the drawings. The following description is an example (representative example) of the present invention, and the present invention is not limited thereto.

[0012] 1 illustrates an ECU 10 according to this embodiment and a common rail system 50 that is the object of its control. The ECU 10 is typically composed of a microcomputer, memories such as RAM and ROM, and the like.

[0013] The common rail system 50 supplies fuel to a diesel engine, which is an internal combustion engine, and includes a fuel tank 51, a low-pressure pump 52, a high-pressure pump 53, a common rail 54, and a plurality of injectors 55.

[0014] The low-pressure pump 52 draws up fuel from the fuel tank 51 and supplies it to the high-pressure pump 53. The high-pressure pump 53 pressurizes the fuel and sends it to the common rail 54. The high-pressure pump 53 is provided with control valves 531 and 532. The control valve 531 adjusts the flow rate of fuel flowing into the high-pressure pump 53. The control valve 532 returns a portion of the fuel to the fuel tank 51, adjusting the pressure of the fuel sent from the low-pressure pump 52 to a constant level. The opening and closing of the control valves 531 and 532 is controlled by the ECU 10.

[0015] The common rail 54 accumulates high-pressure fuel pumped from the high-pressure pump 53 and supplies the high-pressure fuel to each injector 55. The common rail 54 is provided with a sensor 541 and a control valve 542. The sensor 541 detects the pressure of the common rail 54 (hereinafter referred to as rail pressure). The control valve 542 adjusts the rail pressure by returning a portion of the fuel to the fuel tank 51. The opening and closing of the control valve 542 is controlled by the ECU 10.

[0016] The injector 55 is composed of a nozzle body provided with an injection hole, a nozzle needle that moves inside the nozzle body, a control valve 551 that controls the movement of the nozzle needle, etc. The control valve 551 controls the movement of the nozzle needle to open and close the injection hole, and fuel is injected into the cylinder of the diesel engine from the injector 55. The injector 55 may be, for example, an electromagnetically controlled injector that uses an electromagnetic solenoid as the control valve 551, or an electrostrictive injector that uses a piezoelectric actuator as the control valve 551.

[0017] Signals detected by various sensors d1 to dn as well as the rail pressure sensor 541 are input to the ECU 10. The detected signals include parameters of the operating conditions of the engine and common rail system 50, such as the rotation speed of the diesel engine, the temperature of the cooling water for the diesel engine, the temperature of the fuel supplied to the common rail 54, and the accelerator opening of the vehicle.

[0018] The ECU 10 controls the opening and closing of the control valves 542, 551, etc. based on the detection signals, thereby controlling the amount of fuel injected in the common rail system 50. The ECU 10 includes an isochronous control unit 11, a correction unit 12, a common rail control unit 13, and a storage device 14 for controlling the amount of fuel injected.

[0019] The isochronous control unit 11 calculates a target value Y of the amount of fuel injection (hereinafter referred to as the target injection amount) to be supplied from the common rail system 50 by isochronous control. In isochronous control, the ECU 10 normally calculates a target engine speed according to the vehicle's driving mode, and adjusts the required torque, i.e., the target injection amount, to maintain that speed. With isochronous control, the target injection amount is adjusted in accordance with fluctuations in the load on the engine, making it possible to maintain a constant engine speed regardless of the load.

[0020] Such isochronous control has conventionally been used as one of the engine rotation controls in OHW vehicles such as agricultural vehicles, construction vehicles, etc. Since isochronous control itself is not unique to the present invention, the target injection amount Y can be calculated by a known method.

[0021] The correction unit 12 corrects the target injection amount Y calculated by the isochronous control unit 11 and outputs the corrected target injection amount Ya. If the fuel injection amount of the injector 55 fluctuates due to deterioration over time, the target injection amount Y calculated by isochronous control also fluctuates from when the injector 55 was new. The correction unit 12 corrects the target injection amount Y so that the actual injection amount of fuel after deterioration is the same as when the injector was new.

[0022] The common rail control unit 13 generates a control signal to be sent to the control valve 551 of the injector 55 based on the corrected target injection amount Ya. For example, if the control valve 551 is an electromagnetic solenoid, the control signal is a pulse signal. The common rail control unit 13 determines the energization time of the control valve 551 of the injector 55, i.e., the time for injecting fuel, based on the current rail pressure Rp of the common rail 54 and the target injection amount Ya. The common rail control unit 13 generates a pulse signal with a duty ratio according to the determined energization time. The common rail control unit 13 also calculates the target rail pressure of the common rail 54 and generates a control signal to each control valve 531, 532, or 542 of the common rail system 50 so as to achieve the target rail pressure.

[0023] Until isochronous control is started, normal engine rotation control is performed in the common rail control unit 13. That is, target values ​​required for engine rotation control, such as a target injection amount and a target rail pressure, are calculated based on detection signals of the actual engine rotation speed, accelerator opening, rail pressure, etc. The fuel injection amount is controlled based on the calculated target values, and a desired engine rotation state is achieved.

[0024] The storage device 14 stores data such as various maps used for calculations by the isochronous control unit 11, the correction unit 12, and the common rail control unit 13. As the storage device 14, for example, a memory such as a hard disk or a ROM can be used.

[0025] FIG. 2 shows the configuration of the correction unit 12. The correction unit 12 includes an acquisition unit 21, a prediction unit 22, a correction amount calculation unit 231, a storage unit 232, a subtraction unit 233, a multiplication unit 234, an update unit 235, a subtraction unit 241, a calculation unit 242, a subtraction unit 243, a correction value calculation unit 25, and a detection unit 26.

[0026] The correction unit 12 having such a configuration may be realized using hardware resources such as an arithmetic module that performs addition, subtraction, or multiplication, a memory, etc. Alternatively, the correction unit 12 may be realized as software processing by having a processor such as a microcomputer or a CPU read and execute a flowchart described below.

[0027] The acquisition unit 21 acquires the current target injection amount Y calculated by the isochronous control unit 11. The acquisition unit 21 also acquires parameters x1 to xn of the operating conditions of the common rail system 50 required for correcting the target injection amount Y. The parameters x1 to xn include the current rail pressure Rp of the common rail 54 detected by the sensor 541.

[0028] The prediction unit 22 predicts the target injection amount calculated by the isochronous control unit 11 when the common rail system 50 is new, from the parameters x1 to xn. The prediction unit 22 outputs the predicted target injection amount (hereinafter referred to as predicted injection amount) Yp. In this embodiment, the prediction unit 22 has a learning model 22M that performs machine learning such as a neural network, and calculates the predicted injection amount Yp using the learning model 22M.

[0029] The correction amount calculation unit 231 reads out the correction map 23T stored in the storage unit 232, and calculates the correction amount Qc corresponding to the predicted injection amount Yp and the current rail pressure Rp based on the correction map 23T.

[0030] The memory unit 232 stores the correction map 23T. In the correction map 23T, the correction amount Qc for the predicted injection amount Yp is associated for each rail pressure Rp.

[0031] The subtraction unit 233 subtracts the predicted injection amount Yp from the current target injection amount Y, and outputs the difference ΔQy (ΔQy = Y - Yp). The multiplication unit 234 multiplies the difference ΔQy by the learning rate Ir, and outputs the multiplication value Ir×ΔQ. The update unit 235 uses the multiplication value Ir×ΔQ to calculate an updated value of the correction amount Qc associated with the predicted injection amount Yp and the rail pressure Rp, and rewrites the correction amount Qc in the correction map 23T with the updated value.

[0032] The subtraction unit 241 subtracts a preset upper limit value Qlim for the target injection amount at the time of new product from the predicted injection amount Yp, and outputs the difference ΔQlim (ΔQlim = Yp - Qlim). The calculation unit 242 calculates max{ΔQlim, 0}, and outputs the larger one between the difference ΔQlim and 0. The subtraction unit 243 subtracts ΔQlim or 0 output from the calculation unit 242 from the correction amount Qc calculated by the correction amount calculation unit 231, and outputs the corrected correction amount Qc.

[0033] The correction value calculation unit 25 adds the correction amount Qc corrected by the subtraction unit 243 to the current target injection amount Y, and outputs the corrected target injection amount Ya.

[0034] The detection unit 26 determines whether the correction amount Qc output from the correction amount calculation unit 231 is within a certain range of Th1 or more and Th2 or less, that is, whether Th1 ≤ Qc ≤ Th2 is satisfied. Th1 and Th2 are the lower limit value and the upper limit value of the normal correction amount Qc, respectively, and Th1 < Th2 is satisfied. When the correction amount Qc is not within the certain range, the detection unit 26 outputs an error detection signal Error.

[0035] FIG. 3 shows the flow of control of the fuel injection amount in the ECU 10. First, the common rail control unit 13 determines whether the control mode of the fuel injection amount is the isochronous control mode (step S1). If it is not the isochronous control mode (step S1: NO), this process ends. If it is the isochronous control mode (step S1: YES), the common rail control unit 13 enables the correction of the target injection amount Y by the correction unit 12.

[0036] In the correction unit 12, the acquisition unit 21 acquires parameters x1 to xn that represent the current operating conditions of the common rail system 50 and the current target injection amount Y calculated by the isochronous control unit 11 (step S2). The acquisition unit 21 outputs the parameters x1 to xn to the prediction unit 22, and outputs the current target injection amount Y to the subtraction unit 233 and the correction value calculation unit 25. The acquisition unit 21 also extracts a parameter of the current rail pressure Rp from the parameters x1 to xn, and outputs it to the correction amount calculation unit 231 and the update unit 235.

[0037] The prediction unit 22 predicts the target injection amount calculated by isochronous control when the common rail system 50 is new, using the learning model 22M. The prediction unit 22 outputs the predicted injection amount Yp to the correction amount calculation unit 231 and the subtraction units 233 and 241 (step S3). The learning model 22M outputs the predicted injection amount Yp in response to the input of parameters x1 to xn.

[0038] The learning model 22M is constructed in advance by machine learning a plurality of training data. The training data is a data set in which the target injection amount Y calculated by isochronous control when a brand new common rail system 50 is operated is associated with the operating conditions at that time, i.e., the parameters x1 to xn.

[0039] For example, when training the learning model 22M using a neural network, parameters x1 to xn of the training data are input to the learning model 22M. Parameters such as weights and biases in the neural network are updated so that the error between the output from the learning model 22M and the target injection amount Y of the training data is reduced. By repeating the update using multiple pieces of training data, a learning model 22M is constructed that outputs a predicted injection amount Yp when a brand new common rail system 50 is used in response to the input of any parameters x1 to xn.

[0040] Any parameters may be used as the parameters x1 to xn as long as the required fuel injection amount for the engine under load can be calculated by isochronous control. Among different combinations of parameters, a combination that provides high prediction accuracy of the learning model 22M may be selected as the parameters x1 to xn.

[0041] Examples of the various parameters x1 to xn include the following. x1: Vehicle hydraulic oil pressure x2: Vehicle hydraulic oil temperature x3: Common rail pressure x4: Engine RPM x5: Fuel flow rate indication from high pressure pump to common rail x6: Engine coolant temperature x7: Temperature of fuel sent to the common rail x8: Engine oil temperature ... xn: Injector injection pattern

[0042] The parameters x1 to xn can be obtained by detecting the operating conditions of a brand new common rail system 50 using various sensors d1 to dn and 541 in various environments such as high temperature, low temperature, high altitude, or low altitude.

[0043] Common rail systems 50 vary from one another, and the actual injection amount relative to the target injection amount Y may deviate from the standard injection amount, resulting in variations in injection amount between common rail systems 50 of the same type. Therefore, it is preferable to generate training data under various environments by acquiring the parameters x1 to xn of the operating conditions and the target injection amount Y when a new common rail system 50 having a median injection amount characteristic is operated. The median injection amount characteristic means that the actual injection amount relative to the target injection amount Y is within a certain range (standard value ±α) from the standard value. By machine learning such training data, the learning model 22M can output a predicted injection amount Yp when the common rail system 50 is new and has the median injection amount characteristic (hereinafter, a common rail system 50 having the median injection amount characteristic may be referred to as a median product).

[0044] The predicted injection amount Yp is also output to the common rail control unit 13. The common rail control unit 13 calculates the target rail pressure of the common rail 54 from the predicted injection amount Yp based on a rail pressure map. The rail pressure map is a table in which the required target rail pressure is associated with the target injection amount, and is stored in the storage device 14.

[0045] The correction amount calculation unit 231 reads out the correction map 23T from the storage unit 232. The correction amount calculation unit 213 calculates the correction amount Qc based on the correction map 23T from the current rail pressure Rp output from the acquisition unit 21 and the predicted injection amount Yp output from the prediction unit 22 (step S4).

[0046] FIG. 4A shows an example of the correction map 23T. In the correction map 23T, the correction amount Qc [mg] for the predicted injection amount Yp [mg] is associated with each rail pressure Rp [bar]. The correction map 23T can be created by associating the correction amount Qc calculated for the target injection amount when the engine is new and the actual rail pressure Rp (rail pressure detected by the sensor 541) controlled based on the target injection amount. Therefore, the correction amount Qc corresponding to the rail pressure Rp when the engine is new can be obtained from the correction map 23T. The predicted injection amount Yp and the rail pressure Rp in the correction map 23T are grid points represented by coordinates (Yp, Rp). The correction amount Qc between the grid points can be calculated by linearly interpolating the correction amount Qc at each grid point.

[0047] For example, if the predicted injection amount Yp is 10 and the current rail pressure Rp is 600, the correction amount calculation unit 231 acquires the corresponding correction amount Qc of "1.5" from the correction map 23T. If the predicted injection amount Yp is 15 and the current rail pressure Rp is 800, the correction amount calculation unit 231 acquires the correction amounts Qc of the four surrounding grid points, that is, the grid points with coordinates (Yp, Rp) of (10,600), (10,900), (20,600), and (20,900), from the correction map 23T. The correction amount calculation unit 231 calculates the correction amount Qc at the point of coordinate (15,800) by linearly interpolating the correction amount Qc associated with each grid point according to the distance from each grid point to the point of coordinate (15,800).

[0048] Next, subtraction unit 241 subtracts upper limit value Qlim from predicted injection amount Yp and outputs the difference ΔQlim. Upper limit value Qlim is a threshold value that is predetermined for the target injection amount when the engine is new, due to excessive torque limitation. Calculation unit 242 calculates Max{0, ΔQlim} and outputs the larger of 0 and ΔQlim. Next, subtraction unit 243 subtracts ΔQlim from correction amount Qc calculated by correction amount calculation unit 231 (step S5).

[0049] The calculation unit 242 outputs ΔQlim>0 when the predicted injection amount Yp exceeds the upper limit Qlim. Therefore, the correction amount Qc is corrected so that the injection amount is reduced by the excess ΔQlim. This makes it possible to make the target injection amount Ya smaller than the upper limit Qlim, thereby suppressing an excessive increase in torque due to the corrected target injection amount Ya.

[0050] The correction value calculation unit 25 adds the correction amount Qc corrected by the subtraction unit 243 to the current target injection amount Y, thereby outputting a corrected target injection amount Ya (step S6). Based on this target injection amount Ya, the common rail control unit 13 calculates the current conduction time of the injector 55, that is, the time for injecting fuel (step S7). The common rail control unit 13 generates a control signal to the control valve 551 of the injector 55 based on the calculated current conduction time.

[0051] Next, the subtraction unit 233 subtracts the predicted injection amount Yp from the current target injection amount Y and outputs the difference ΔQy (step S8). The multiplication unit 234 multiplies the difference ΔQy by a learning rate Ir (step S9). The learning rate Ir is a coefficient used to update the correction map 23T. The learning rate Ir can be set arbitrarily, but is typically 0.2 to 0.5. The update unit 235 updates the correction amount Qc of the correction map 23T according to the multiplication value Ir×ΔQy and the current rail pressure Rp (step S10).

[0052] First, the update unit 235 acquires the correction amount Qc corresponding to the predicted injection amount Yp and the current rail pressure Rp from the correction map 23T. The update unit 235 calculates the updated correction amount Qc by adding the multiplication value Ir×ΔQy to this correction amount Qc. If the predicted injection amount Yp and the current rail pressure Rp are between grid points, the update unit 235 updates the correction amounts Qc of the surrounding grid points according to the distance from the surrounding grid points.

[0053] For example, if the predicted injection amount Yp is 10, the current rail pressure Rp is 800, the difference ΔQy is 0.3, and the learning rate Ir is 0.5, the update unit 235 updates the correction amount Qc at two lattice points (10,600) and (10,900) surrounding the coordinate (10,800). Since the correction amounts Qc at the respective lattice points in the correction map 23T are 1.5 and 1.2, the update unit 235 calculates the updated values ​​of these correction amounts Qc according to the distance from the coordinate (10,800) to each lattice point. The following formula is the calculation formula. 1.5 + {1 - (800 - 600) / (900 - 600)} x 0.3 x 0.5 = 1.55 1.2 + {1 - (900 - 800) / (900 - 600)} x 0.3 x 0.5 = 1.3

[0054] FIG. 4B shows the updated correction map 23T. The update unit 235 rewrites the correction amount Qc of the lattice point (10,600) in the correction map 23T from 1.5 to an updated value of 1.55. The update unit 235 also rewrites the correction amount Qc of the lattice point (10,900) from 1.2 to an updated value of 1.3.

[0055] The initial correction amount Qc used to create the correction map 23T is the difference ΔQy obtained by subtracting the predicted injection amount Yp when the common rail system 50 is new and in the average condition from the target injection amount Y calculated when the common rail system 50 is new. This difference ΔQy represents the variation in the injection amount from the standard value. At the time of the initial correction, the difference ΔQy corresponding to the variation is multiplied by the learning rate Ir and added to the target injection amount Y as the correction amount Qc to calculate the corrected target injection amount Ya. The corrected target injection amount Ya is used to calculate the energization time of the injector 55 and inject fuel, so the actual injection amount is also corrected.

[0056] During the second correction, the target injection amount Y is calculated by isochronous control based on the engine torque generated by the actual injection amount corrected during the first correction. Therefore, the difference ΔQy between the target injection amount Y and the predicted injection amount Yp becomes smaller compared to the first correction. The correction map 23T is then updated using the difference ΔQy multiplied by the learning rate Ir. The correction amount Qc output from the updated correction map 23T is added to the second target injection amount Y, and the second corrected target injection amount Ya is calculated. This cycle is then repeated, and the correction amount Qc is successively updated according to the difference ΔQy between the target injection amount Y for the aged common rail system 50 and the predicted injection amount Yp for a new product. As a result, the difference (Yp - Y) in the target injection amount from a new, median product gradually becomes smaller.

[0057] FIG. 5 shows the correlation between the target injection amount Y, the predicted injection amount Yp, and the corrected target injection amount Ya. In the case of a common rail system 50 in which the amount of fuel injected is smaller than that of a median product when new and the actual injection amount gradually decreases due to deterioration over time, the calculated target injection amount Y gradually becomes smaller than the corrected target injection amount Ya as time t passes.

[0058] By accumulating this difference (Yp-Y) as ΔQy in the correction map 23T, the difference (Yp-Y) gradually decreases, and the current target injection amount Y comes to match the predicted injection amount Yp. The time until they match can be adjusted by setting the learning rate Ir. Even if the actual injection amount decreases due to deterioration, learning is carried out successively to fill the difference between the target injection amount Y and the predicted injection amount Yp, so the correction amount Qc is increased according to the amount of decrease in the injection amount, and the target injection amount Ya is calculated so that the injection amount is always the same as that of a new, median product.

[0059] On the other hand, the detection unit 26 determines whether the correction amount Qc output from the correction amount calculation unit 231 is within a certain range from Th1 to Th2 (step S11). If it is not within the certain range (step S11: NO), the current target injection amount Y is significantly different from the predicted injection amount Yp, and there is a possibility that an abnormality has occurred. Therefore, in this case, the detection unit 26 outputs an error detection signal Error (step S12).

[0060] When the detection signal Error is output, the common rail control unit 13 may notify the driver of the possibility of an abnormality by, for example, displaying an abnormality notification on a display.

[0061] As described above, according to this embodiment, the prediction unit 22 predicts the target injection amount calculated by isochronous control when the common rail system 50 is new, based on the parameters x1 to xn that represent the current operating conditions of the common rail system 50, and outputs the predicted injection amount Yp when the common rail system 50 is new. The correction amount calculation unit 231 calculates the correction amount Qc that corresponds to the predicted injection amount Yp, based on the correction map 23T. The correction value calculation unit 25 adds the correction amount Qc to the current target injection amount Y calculated by isochronous control, and outputs the corrected target injection amount Ya.

[0062] The current target injection amount Y is corrected by adding the correction amount Qc so that the difference with the predicted injection amount Yp is eliminated, and the actual fuel injection amount achieved by the current conduction time calculated using the corrected target injection amount Ya is the same as or close to that of a new engine. This reduces fluctuations in the actual fuel injection amount due to deterioration over time.

[0063] Conventionally, an upper limit for torque limitation was set for the target injection amount Y calculated by isochronous control, so when the target injection amount Y increases and reaches the upper limit, the target injection amount Y cannot be increased any further, and high-load work similar to that when the engine was new is restricted. However, according to this embodiment, the target injection amount Y is corrected so as to approach the target injection amount when the engine was new, so such torque limitation due to deterioration over time is not imposed.

[0064] On the other hand, if the predicted injection amount Yp exceeds the upper limit Qlim of the target injection amount for a new engine, the excess amount is subtracted from the predicted injection amount Yp, so the corrected target injection amount Ya does not exceed the upper limit for a new engine. Therefore, high-load work similar to that of a new engine can be performed while suppressing torque increases above the upper limit.

[0065] Because the target value of rail pressure Rp is always calculated from the predicted injection amount Yp, rail pressure Rp remains unchanged from the state it was in when it was new. Even if the target injection amount Y calculated by isochronous control increases as the actual injection amount decreases due to aging, the accompanying change in emissions can be avoided.

[0066] The learning model 22M of the prediction unit 22 learns training data acquired from multiple brand new common rail systems 50, and is therefore able to predict the predicted injection amount Yp of a brand new, median product. This makes it possible to correct for variations in injection amount due to individual differences between each common rail system 50. When used in conjunction with correction of variations between engine cylinders, the actual injection amount of each cylinder can be corrected to a level close to that of a brand new, median product, even after deterioration.

[0067] Although the preferred embodiments of the present invention have been described above, the present invention is not limited to these embodiments and various modifications are possible within the scope of the present invention.

[0068] For example, a recording medium may be provided that stores a program for causing a computer to execute the control method of the present invention. The recording medium is not particularly limited as long as it is a recording medium that can be read by a computer such as a CPU, and semiconductor memory, magnetic disk, optical disk, or the like can be used. [Explanation of symbols]

[0069] 10···ECU, 11···Isochronous control unit, 12···Correction unit, 13···Common rail control unit, 22···Prediction unit, 22M···Learning model, 231···Correction amount calculation unit, 232···Memory unit, 23T···Correction map, 25···Correction value calculation unit, 26···Detection unit

Claims

1. A control device (10) that adjusts the amount of fuel injected in a common rail system (50) by isochronous control, a prediction unit (22) that predicts a target injection amount calculated by the isochronous control when the common rail system (50) is new based on current operating conditions (x1 to xn) of the common rail system (50), and outputs a predicted injection amount (Yp) when the common rail system (50) is new; a storage unit (232) for storing a correction map (23T) in which a correction amount (Qc) is associated with the predicted injection amount (Yp); a correction amount calculation unit (231) that calculates the correction amount (Qc) corresponding to the predicted injection amount (Yp) based on the correction map (23T); a correction value calculation unit (25) that outputs a corrected target injection amount (Ya) by adding the correction amount (Qc) to a current target injection amount (Y) calculated by the isochronous control; Equipped with In the correction map (23T), the correction amount (Qc) with respect to the predicted injection amount (Yp) is associated with each rail pressure (Rp) of the common rail system (50), The correction amount calculation unit (231) calculating the correction amount (Qc) corresponding to the current rail pressure (Rp) and the predicted injection amount (Yp) based on the correction map (23T); The control device further an updating unit (235) that updates the correction amount (Qc) of the correction map (23T) based on a difference (ΔQy) between the predicted injection amount (Yp) and the current target injection amount (Y); The prediction unit (22) a learning model (22M) that receives current operating conditions (x1 to xn) of the common rail system (50) and outputs the predicted injection amount (Yp) under the input current operating conditions (x1 to xn) of the common rail system (50); The learning model (22M) A control device (10) that performs machine learning in advance training data that associates operating conditions (x1 to xn) of the common rail system (50) when new with a target injection amount (Y) calculated by the isochronous control under the operating conditions (x1 to xn) of the common rail system (50) when new.

2. A subtraction unit (233) that calculates the difference (ΔQy); a multiplication unit (234) that calculates the product of the difference (ΔQy) and a learning rate (Ir); Further provided with The update unit (235) The correction amount (Qc) is updated by adding the multiplied value to the correction amount (Qc) of the correction map (23T). The control device (10) of claim 1.

3. A detection unit (26) is provided to detect an error when the correction amount (Qc) is not within a certain range. A control device (10) according to claim 1 or 2.

4. a subtraction unit (241) that calculates a difference (ΔQlim) between the predicted injection amount (Yp) and an upper limit value (Qlim) determined for a target injection amount for a new product; a subtraction unit (243) that corrects the correction amount (Qc) by subtracting the difference (ΔQlim) from the correction amount (Qc) when the difference (ΔQlim) is greater than 0; The correction value calculation unit (25) adds the corrected correction amount (Qc) to the current target injection amount (Y). A control device (10) according to any one of claims 1 to 3.

5. The learning model (22M) a plurality of training data in which operating conditions (x1 to xn) when the common rail system (50) that is new and has a median characteristic of the injection amount are operated are associated with a target injection amount (Y) under the operating conditions, and In response to input of the current operating conditions (x1 to xn) of the common rail system (50), the predicted injection amount (Yp) is output when the common rail system (50) is new and has a central injection amount characteristic. A control device (10) according to any one of claims 1 to 4.

6. A control method for adjusting a fuel injection amount in a common rail system (50) by isochronous control, comprising: a step of predicting a target injection amount calculated by the isochronous control when the common rail system (50) is new based on current operating conditions (x1 to xn) of the common rail system (50), and outputting a predicted injection amount (Yp); calculating a correction amount (Qc) corresponding to the predicted injection amount (Yp) based on a correction map (23T) in which the correction amount (Qc) is associated with the predicted injection amount (Yp); a step of adding the correction amount (Qc) to the current target injection amount (Y) calculated by the isochronous control, thereby outputting a corrected target injection amount (Ya); Including, In the correction map (23T), the correction amount (Qc) with respect to the predicted injection amount (Yp) is associated with each rail pressure (Rp) of the common rail system (50), The correction amount (Qc) is associated with the current rail pressure (Rp) and the predicted injection amount (Yp) in the correction map (23T), The control method further comprises: updating the correction amount (Qc) of the correction map (23T) based on a difference (ΔQy) between the predicted injection amount (Yp) and the current target injection amount (Y); The predicted injection amount (Yp) is The predicted injection amount (Yp) under the current operating conditions (x1 to xn) of the common rail system (50) obtained from a learning model (22M), The learning model (22M) is A control method for machine learning training data in advance that associates operating conditions (x1 to xn) of the common rail system (50) when new with a target injection amount (Y) calculated by the isochronous control under the operating conditions (x1 to xn) of the common rail system (50) when new.

Images