JP2014234923A - Solenoid valve driving controller - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a solenoid valve drive controller capable of efficiently doing learning necessary for solenoid valve opening time control and achieving calculation load reduction and learning time reduction.SOLUTION: Valve opening command time Ti for a fuel injection valve 2 is calculated on the basis of valve opening request time Topen, and the fuel injection valve 2 is driven using the valve opening command time Ti. When the fuel injection valve 2 is driven, actual valve opening time TopenA is detected and the valve opening command time Ti is corrected such that the actual valve opening time TopenA is closer to the valve opening request time Topen. In addition, a Ti table is updated using the corrected valve opening command time Ti and the fuel injection valve 2 is driven using the updated Ti table. Within an interpolation learning region RINTLRN, a preceding learning is conducted at two preceding learning points, thereby performing an initial setting of a Ti table set value at each learning point other than the preceding learning points.

Description

  The present invention relates to a drive control device for a solenoid valve, and more particularly to a device that controls opening and closing of a solenoid valve that controls the flow rate of a fluid, such as a fuel injection valve and an exhaust gas recirculation control valve mounted on an internal combustion engine.

  Patent Document 1 discloses a fuel injection device including a fuel injection valve provided in an internal combustion engine. According to this device, the actual lift period of the fuel injection valve is detected using the lift sensor, and the drive signal of the fuel injection valve is corrected so that the actual lift period coincides with the target lift period. Further, it describes that a function for learning the correction result is added.

JP 63-97869 A

  For example, in a fuel injection valve that directly injects fuel into the combustion chamber of an internal combustion engine, a valve opening command time that is set so that the actual valve opening time (actual lift period) of the fuel injection valve matches the valve opening request time, Since the relationship with the required valve time changes depending on the environmental conditions such as characteristic variation and fuel pressure during mass production, it is required to reflect this in the control while learning this relationship.

  In the above Patent Document 1, there is no detailed description of the learning function, and the calculation load of the control device and the time required for learning are improved when learning the optimum relationship between the valve opening command time and the valve opening request time. There was room for.

  The present invention has been made paying attention to this point, and can efficiently perform learning necessary for valve opening time control of the solenoid valve, and reduce the calculation load and the learning time. The purpose is to provide.

  In order to achieve the above object, according to the first aspect of the present invention, in the drive control device for the electromagnetic valve (2) for controlling the flow rate of the fluid, the valve opening request time for calculating the valve opening request time (Topen). The valve opening command time (Ti) of the solenoid valve is calculated based on the calculation means and the valve opening request time (Topen), and the drive signal of the solenoid valve (2) is calculated using the valve opening command time (Ti). Drive signal generation means for generating the actual valve opening time detection means (TopenA) for detecting the actual valve opening time (TopenA) when the solenoid valve is driven, and the actual valve opening time (TopenA) is the valve opening request. The valve opening command time (Ti) is corrected so as to approach the time (Topen), and learning value holding means for holding the corrected valve opening command time as a learning value is provided. Previous using learning value A drive signal is generated, and the learning value holding means includes two learning points corresponding to a lower valve opening request time and an upper valve opening request time that divide the learning region, and is set in advance in the learning region (RINTLRN) The preceding learning point (PL) is obtained by performing the preceding learning at each of the number of preceding learning points (PL, PH) smaller than the number of all learning points to be performed and performing interpolation calculation of the preceding learning value obtained by the preceding learning. , PH), initial values of the learning values are set at learning points.

  According to this configuration, the valve opening command time of the solenoid valve is calculated based on the valve opening request time, and a drive signal for the solenoid valve is generated using the valve opening command time. When the solenoid valve is driven, the actual valve opening time is detected, and the valve opening command time is corrected so that the actual valve opening time approaches the valve opening request time. Retained. A drive signal is generated using the stored learning value. Pre-learning is performed at each of the pre-learning points less than the total number of learning points set in advance in the learning area, and learning other than the pre-learning points is performed by performing interpolation calculation of the pre-learning value obtained by the pre-learning The initial value of the learning value at the point is set. Therefore, for example, if the preceding learning points are two points corresponding to the lower valve opening request time and the upper valve opening request time, compared to the case where the preceding learning for initial value setting is performed at all the learning points in the learning region. The learning time and calculation load required for initial value setting can be greatly reduced.

  According to a second aspect of the present invention, in the electromagnetic valve drive control device according to the first aspect, the learning value holding means starts learning in the learning region (RINTLRN) after setting the initial value. Features.

  According to this configuration, since learning in the learning area is started after setting the initial value, the learning value at the learning point can be quickly converged to the optimum value.

  According to a third aspect of the present invention, in the electromagnetic valve drive control device according to the first or second aspect, the learning region (RINTLRN) includes the valve opening request time (Topen) and the valve opening command time (Ti). The relationship is a region affected by environmental conditions affected by environmental conditions.

  According to this configuration, the learning area is an environmental condition influence area in which the relationship between the valve opening request time and the valve opening command time is affected by the environmental conditions. In the area affected by environmental conditions, the optimal relationship between the valve opening request time and the valve opening command time is close to a linear relationship. Accuracy is obtained.

  According to a fourth aspect of the present invention, in the electromagnetic valve drive control device according to any one of the first to third aspects, the electromagnetic valve (2) includes a core (35) on which an electromagnetic force acts, a valve The valve shaft (31) to which the body (32) is fixed has a hammering core structure configured separately, and the valve opening required time (Topen) is shorter than the environmental condition influence region. The relationship between the required time (Topen) and the valve opening command time (Ti) has a non-linear region (RALPLRN) in which the relationship is non-linear, and the learning value holding means has all the preset values in the non-linear region (RALPLRN). The initial value setting is performed by executing the preceding learning at a learning point.

  According to this configuration, the solenoid valve has a hammering core structure, and has a non-linear region where the relationship between the valve opening request time and the valve opening command time is non-linear on the side where the valve opening request time is shorter than the environmental condition influence region. In the non-linear region, the initial value is set by executing pre-learning at all learning points set in advance. In the non-linear region, if initial value setting is performed by interpolation calculation, sufficient setting accuracy cannot be obtained. Therefore, by performing prior learning at all learning points, relatively high control accuracy can be ensured from the beginning of control. .

1 is a diagram illustrating an internal combustion engine and a control device thereof according to an embodiment of the present invention. It is sectional drawing for demonstrating the structure of the principal part of a fuel injection valve. It is a figure which shows the table applied in order to calculate valve opening command time (Ti) according to valve opening request | requirement time (Topen). It is a time chart which shows the state where the table setting value of valve opening command time (Ti) converges by initial learning. It is a time chart for demonstrating the calculation method of actual valve opening time (TopenA). It is a flowchart of the process which performs initial learning. It is a flowchart of the interpolation calculation process performed by the process of FIG. It is a flowchart of the fuel-injection control process which controls fuel-injection by a fuel-injection valve. It is a figure which shows an example of the table corrected by learning. It is a figure which shows the modification regarding the setting of an interpolation learning area | region (RINTLRN).

Embodiments of the present invention will be described below with reference to the drawings.
FIG. 1 is a view showing an internal combustion engine (hereinafter referred to as an “engine”) and a control device thereof according to an embodiment of the present invention. In this embodiment, a fuel injection valve opened by a solenoid valve having a solenoid is opened. The amount of fuel supplied to the engine is controlled by changing the time.

  The four-cylinder engine 1 includes four fuel injection valves 2 corresponding to the cylinders, and the fuel injection valves 2 directly inject fuel into the combustion chamber of the engine 1. Each of the four fuel injection valves 2 is connected to the ECU 5, and its operation is controlled by the ECU 5.

  The fuel injection valve 2 is connected to a delivery pipe 4 through a fuel passage 3, and fuel pressurized by a high-pressure fuel pump (not shown) is supplied to the delivery pipe 4. A fuel pressure sensor 12 for detecting the fuel pressure PF is attached to the delivery pipe 4, and the detection signal is supplied to the ECU 5.

  The ECU 5 includes a voltage / current detection sensor 11 for detecting the voltage VSL across the solenoid of the fuel injection valve 2 and the drive current ID supplied to the solenoid, an engine speed sensor 13 for detecting the engine speed NE, and the engine 1. Various sensors for detecting the engine operating state are connected, such as an intake air flow rate sensor 14 for detecting the intake air flow rate GAIR, an intake air temperature sensor 15 for detecting the intake air temperature TA, and a cooling water temperature sensor 16 for detecting the engine cooling water temperature TW. The detection signals of these sensors are supplied to the ECU 5. The ECU 5 calculates the valve opening request time Topen of the fuel injection valve 2 according to the engine operating state using the detection signals of these sensors, calculates the valve opening command time Ti according to the valve opening request time Topen, and opens the valve. Drive control of the fuel injection valve 2 is performed using the valve command time Ti.

  FIG. 2 is a cross-sectional view for explaining a configuration of a main part of the fuel injection valve 2. The fuel injection valve 2 includes a valve shaft 31, a valve body 32 fixed to the tip of the valve shaft 31, and a valve shaft 31. Flanges 33, 34 fixed to the core, a core 35 on which electromagnetic force acts, a first spring 36 provided between the core 35 and the flange 34, a valve seat 37, a sleeve 38, a solenoid 39, A second spring 40 that urges the flange 34 in the valve closing direction (downward in the figure) and an inner collar 41 having a hollow portion that functions as a fuel passage are provided. The fuel injection valve 2 has a so-called hammering core structure in which a core 35 and a valve shaft 31 to which a valve body 32 is fixed are configured separately.

  FIG. 3 is a diagram showing a relationship (Ti table) between the valve opening request time Topen and the valve opening command time Ti applied to calculate the valve opening command time Ti according to the valve opening request time Topen. The solid line L1 shown in this figure indicates the relationship for setting the start value when performing prior learning of the valve opening command time Ti corresponding to the valve opening request time Topen, and the solid line L2 indicates the actual valve opening time TopenA. Shows the relationship between the valve opening command time Ti set so as to coincide with the valve opening request time Topen and the valve opening request time Topen.

  In this embodiment, initial values are set by initial learning of the Ti table as follows according to the valve opening request time Topen. Initial learning is performed at the time of factory shipment or when the fuel injection valve 2 is replaced with a new one, and thereafter, the table setting values are updated by learning during actual operation of the engine 1.

  1) In a region where the valve opening request time Topen is larger than the upper valve opening request time TOLH (for example, 1.0 msec), the valve opening command time Ti is calculated using a relationship (not shown) indicated by a straight line obtained in advance. To do. That is, since the table setting value is not learned, the initial value setting (initial learning) is unnecessary.

  2) In a region where the valve opening request time Topen is equal to or shorter than the upper valve opening request time TOLH and is longer than the lower valve opening request time TOLL (for example, 0.5 msec) (hereinafter referred to as “interpolation learning region RINTLRN”), the solid line L2 is relatively straight. Focusing on the fact that the relationship is close to, the initial value is set by simplified initial learning (interpolation learning). In the interpolation learning region RINTLRN, fuel is actually used with the valve opening command time Ti corresponding to the learning start points PHI and PLI on the solid line L1 corresponding to the upper valve opening request time TOLH and the lower valve opening request time TOLL as a start value. Pre-learning is started by detecting the actual valve opening time TopenA by performing injection, and by gradually updating the valve opening command time Ti, the pre-learning points PH and PL are obtained, and other values in the interpolation learning region RINTLRN are obtained. As for the set grid points (learning points), initial values are set by interpolation using a straight line (broken line L3) connecting the preceding learning points PH and PL as shown in FIG.

  3) In a region where the valve opening request time Topen is smaller than the lower valve opening request time TOLL (hereinafter referred to as “all-point learning region RALPLRN”), the preceding learning is executed at all points of the set grid points on the table to set initial values. I do. The set table setting value by the preceding learning is indicated by a broken line L4, for example.

  Note that only the representative learning points indicated by white circles in FIG. 3 are illustrated, and are actually set at smaller intervals (for example, 20 μsec intervals).

  FIG. 4 is a time chart showing a state in which the valve opening command time Ti (table setting value) converges by learning, and FIG. 4A sets the feedback control gain GP to a relatively small value “0.1”. FIG. 4B corresponds to the case where the feedback control gain GP is set to a relatively large value “0.4”. 4 (a) and 4 (b), it can be confirmed that the larger the gain GP, the shorter the time required for convergence, but the greater the fluctuation after convergence. Therefore, it is desirable to set the gain GP to a relatively small value and shorten the total learning time required for the initial learning.

  For example, if there are 100 learning points and the learning time required for convergence at each learning point is 2 seconds, the total learning time is 200 seconds. In the interpolation learning area RINTLRN shown in FIG. Is only 2 points (PH, PL), and initial values are set at other learning points by interpolation, so that the total learning time is 2 seconds + α (a little time required for interpolation), and the gain GP is a small value. The total learning time can be greatly shortened while setting to.

  FIG. 5 is a time chart for explaining the calculation method of the actual valve opening time TopenA. FIG. 5A shows the transition of the valve body lift amount LFT of the fuel injection valve 2, and FIG. Indicates a valve opening command signal. In the present embodiment, the actual valve opening timing tOP is obtained from the inflection point of the current waveform detected by the voltage / current sensor 11, and the actual valve closing timing tCL is the inflection point of the voltage waveform detected by the voltage / current sensor 11. Is required. This detection method is disclosed in, for example, Japanese Patent Laid-Open No. 6-174139.

More specifically, the valve opening delay time Ton from the valve opening command timing tIS to the actual valve opening timing tOP and the valve closing operation time Toff from the valve closing command timing tIE to the actual valve closing timing tCL are calculated, and the valve opening command The actual valve opening timing TopenA is calculated by applying to the following formula (1) together with the time Ti.
TopenA = Ti-Ton + Toff (1)

  FIG. 6 is a flowchart of the initial setting process including the preceding learning described above. This process is executed in the ECU 5. It should be noted that all the calculations related to the calculation of the valve opening command time Ti described below are performed for each cylinder to be controlled.

  In step S11, the valve opening request times Topen [1] and Topen [2] corresponding to the preceding learning point are set to the lower valve opening request time TOLL and the upper valve opening request time TOLH, respectively, and the index parameter i is set to “0”. To "".

  In step S12, the index parameter i is increased by “1”, and the valve opening command time Ti [i] corresponding to the valve opening request time Topen [i] is calculated (step S13). For the calculation in step S13, for example, the valve opening command time Ti [i] is calculated by applying the relationship of the solid line L1 shown in FIG.

  In step S14, fuel injection is performed, and the above-described valve opening delay time Ton and valve closing operation time Toff are detected (step S15). In step S16, the actual valve opening time TopenA is calculated using the above equation (1). In step S17, the valve opening time deviation DTopen is calculated by subtracting the actual valve opening timing TopenA from the valve opening request time Topen.

In step S18, it is determined whether or not the absolute value of the valve opening time deviation DTopen is equal to or less than the convergence determination threshold DLL. Initially, the answer is negative (NO), the process proceeds to step S19, and the valve opening command time Ti [i] is updated using the following equation (2). GP in Equation (2) is a control gain.
Ti [i] = Ti [i] + GP × DTOpen (2)

  After execution of step S19, the process returns to step S14. When the answer to step S18 is affirmative (YES), that is, when the valve opening command time Ti [i] has converged, whether or not the index parameter i is “2”. Is determined (step S20). Since this answer is negative (NO) at first, the process returns to step S12, and the same processing is executed when i = 2, that is, for the valve opening request time Topen [2]. If the answer to step S18 is affirmative (YES), the answer to step S20 is also affirmed (YES) this time, so the process proceeds to step S21 to execute the interpolation calculation process shown in FIG.

  In step S31 of FIG. 7, the parameter x indicating the valve opening request time Topen is set to the lower valve opening request time TOLL, and in step S32, the parameter x is increased by the lattice point interval DX (for example, 20 μsec) of the Ti table. .

In step S33, an initial learning value Ti (x) of the valve opening command time is calculated by an interpolation calculation of the following equation (3).

In step S34, the Ti table is updated by setting the value of the corresponding lattice point in the Ti table to the initial learning value Ti (x). In step S35, it is determined whether or not the parameter x is equal to or longer than the upper valve opening request time TOLH. Initially, this answer is negative (NO), and the process returns to step S32 to repeat the processes of steps S32 to S34. If the answer to step S35 is affirmative (YES), the process ends.
With the processing in FIG. 7, the initial value setting of the lattice points (learning points) corresponding to the white circles indicated by the broken line L3 in FIG. 3 is completed.

  Returning to FIG. 6, in step S <b> 22, the start value of the preceding learning (indicated by a white circle on the solid line L <b> 1 in FIG. 3) at all learning points (DX intervals) in the all-point learning region RALPLRN by the same method as steps S <b> 14 to S <b> 19. Execute pre-learning and set initial values. The set initial value is indicated by a white circle on the broken line L4 in FIG.

  FIG. 8 is a flowchart of the fuel injection control process for controlling the fuel injection by the fuel injection valve 2.

  In step S41, the valve opening request time Topen is calculated according to the engine operating state, and in step S42, the initial value set Ti table (interpolation learning) shown in FIG. 3 is initially set in accordance with the calculated valve opening request time Topen. The area RINTLRN is indicated by a broken line L3, and the all-point learning area RALPLRN is indicated by a broken line L4) to calculate a valve opening command time Ti.

  The processing in steps S43 to S47 is the same as the processing in steps S14 to S19 in FIG. 6, and the actual valve opening time TopenA is calculated when fuel injection is executed, and the actual valve opening time TopenA converges to the required time Topen. As described above, the valve opening command time Ti is corrected.

In step S48, the corrected valve opening command time Ti is applied to the following equation (4) to update (learn) the Ti table. Ti on the right side is the valve opening command time corrected in step S47, TiP is a table setting value before update, and CL is an annealing coefficient set to about 0.1, for example.
Ti = CL * Ti + (1-CL) * TiP (4)

  By the process of FIG. 8, learning of the Ti table (setting value update) is performed while actually performing fuel injection, and the setting value of the Ti table gradually converges to an optimum value for the fuel injection valve 2 being used. To do. That is, in step S42, the valve opening command time Ti is initially calculated using the initial value set by the initial learning process, but the Ti table set value is gradually updated by operating the engine 1, The valve opening command time Ti is calculated using the updated set value. FIG. 9 shows an example of a Ti table corrected by learning.

  As described above, in the present embodiment, the valve opening command time Ti of the fuel injection valve 2 corresponding to the valve opening request time Topen is calculated using the Ti table, and the fuel injection valve 2 is driven using the valve opening command time Ti. Is done. The actual valve opening time TopenA is detected when the fuel injection valve 2 is driven, the valve opening command time Ti is corrected so that the actual valve opening time TopenA approaches the valve opening request time Topen, and the corrected valve opening command time Ti Thus, the Ti table is updated (learned) and held as a table set value, and the fuel injection valve 2 is driven using the held table set value. In the interpolation learning area RINTLRN, the preceding learning is performed at two preceding learning points smaller than the total number of learning points set in advance, that is, at two points corresponding to the lower valve opening request time TOLL and the upper valve opening request time TOLH. At the same time, by performing interpolation calculation of the preceding learning values (values corresponding to PL and PH) obtained by the preceding learning, the initial value setting of the Ti table setting values at the learning points other than the preceding learning points is performed. Therefore, the learning time and calculation load required for the initial value setting can be greatly reduced as compared with the case where the preceding learning for the initial value setting is performed at all the learning points in the interpolation learning region RINTLRN.

Moreover, since learning of the set value in the interpolation learning area RINTLRN is started after the initial value is set, the learned value at the learning point can be quickly converged to the optimum value.
The interpolation learning area RINTLRN is an environmental condition influence area in which the relationship between the valve opening request time Topen and the valve opening command time Ti is affected by environmental conditions, that is, the influence of variation in characteristics and fuel pressure during mass production. In the environmental condition influence region, since the optimum relationship between the valve opening request time Topen and the valve opening command time Ti is close to a linear relationship, sufficient setting accuracy can be obtained by the initial value setting by the interpolation operation, High control accuracy can be obtained.

  Further, the fuel injection valve 2 has a hammering core structure in which a core 35 on which electromagnetic force acts and a valve shaft 31 to which a valve body 32 is fixed are configured separately, and the valve opening request time Topen from the environmental condition influence region. Has a nonlinear region where the relationship between the valve opening request time Topen and the valve opening command time Ti is nonlinear. In the present embodiment, this non-linear area is set as an all-point learning area RALPLRN, and the initial value is set by executing pre-learning at all preset learning points. In the all-point learning region RALPLRN, since sufficient setting accuracy cannot be obtained when initial value setting is performed by interpolation calculation, relatively high control accuracy is secured from the beginning of control by performing advance learning at all learning points. be able to.

  The present invention is not limited to the embodiment described above, and various modifications can be made. For example, the interpolation learning area RINTLRN and the all-point learning area RALPLRN are not limited to the settings shown in FIG. 3. For example, as shown in FIG. 10, the interpolation learning area RINTLRN has a smaller valve opening request time Topen. You may make it expand to the side. This is because even in the interpolation learning region RINTLRN shown in FIG. 10, a certain degree of linearity is ensured in the relationship between the valve opening request time Topen and the valve opening command time Ti.

  In the above-described embodiment, the preceding learning points in the interpolation learning area RINTLRN are two points corresponding to the lower limit value (TOLL) and the upper limit value (TOLH) that divide the interpolation learning area RINTLRN. However, the present invention is not limited to this. Alternatively, initial setting by preceding learning may be performed at preceding learning points smaller than the total number of learning points, and initial value setting at learning points other than the preceding learning points may be performed by interpolation calculation.

In the embodiment described above, an example in which the present invention is applied to a fuel injection valve of an internal combustion engine has been described. However, the present invention can also be applied to a general electromagnetic valve for controlling the flow rate of a fluid.
Further, the actual valve opening timing tOP and the actual valve closing timing tCL may be detected using a lift sensor as disclosed in Patent Document 1.

2 Fuel injection valve (solenoid valve)
5 Electronic control unit (valve opening request time calculating means, actual valve opening time detecting means, drive signal generating means, learning value holding means)
11 Voltage current sensor (actual valve opening time detection means)
31 Valve shaft 32 Valve body

Claims (4)

In a drive control device for a solenoid valve that controls the flow rate of fluid,
Valve opening request time calculating means for calculating the valve opening request time of the solenoid valve;
Drive signal generating means for calculating a valve opening command time of the solenoid valve based on the valve opening request time, and generating a drive signal of the solenoid valve using the valve opening command time;
An actual valve opening time detecting means for detecting an actual valve opening time of the electromagnetic valve when the electromagnetic valve is driven;
Learning value holding means for correcting the valve opening command time so that the actual valve opening time approaches the valve opening request time, and holding the corrected valve opening command time as a learning value;
The drive signal generation means generates the drive signal using the learning value,
The learning value holding means includes two learning points corresponding to the lower valve opening request time and the upper valve opening request time that divide the learning region, and is smaller than the total number of learning points set in advance in the learning region. Performing preceding learning at each of a number of preceding learning points and performing initial calculation of the learning values at learning points other than the preceding learning points by performing interpolation calculation of the preceding learning values obtained by the preceding learning. A drive control device for a solenoid valve.   The electromagnetic valve drive control device according to claim 1, wherein the learning value holding means starts learning in the learning region after the initial value is set.   The electromagnetic valve drive control device according to claim 1, wherein the learning region is an environmental condition influence region in which a relationship between the valve opening request time and the valve opening command time is affected by an environmental condition. . The solenoid valve has a hammering core structure in which a core on which an electromagnetic force acts and a valve shaft to which the valve body is fixed are configured separately,
On the side where the valve opening request time is shorter than the environmental condition influence region, there is a non-linear region where the relationship between the valve opening request time and the valve opening command time is non-linear,
4. The drive control device for an electromagnetic valve according to claim 3, wherein the learning value holding unit performs the preceding learning at all learning points set in advance in the non-linear region to set the initial value. 5. .

Images