WO2006081876A1

WO2006081876A1 - Method for equalising a sensor signal, in particular the input signal of a control device for an internal combustion engine

Info

Publication number: WO2006081876A1
Application number: PCT/EP2005/052766
Authority: WO
Inventors: Andreas Pflüger
Original assignee: Siemens Aktiengesellschaft
Priority date: 2004-07-12
Filing date: 2005-06-15
Publication date: 2006-08-10
Also published as: DE102004033615B3

Abstract

The invention relates to a method for equalising a sensor signal, in particular the input signal of a control device for an internal combustion engine consisting in filtering said sensor signal with the aid of a high pass filter, in determining the time constant of said high pass filter using the actual value of the output signal thereof and in filtering the sensor signal by means of the high pass filter provided with a new time constant.

Description

description

Method for smoothing a sensor signal, in particular an input signal of a control unit of a Brennkraftraa- machine

The invention relates to a method for smoothing a sensor signal, in particular an input signal of a control unit of an internal combustion engine, with a first-order low-pass filter.

In the control and regulation of internal combustion engines, electronic engine control units are used which read and process a wide variety of sensor signals and calculate therefrom the control signals for adjusting the fuel injection quantity and, if no self-igniting engine is present, the ignition timing. The processing of the sensor signals can take place in an analogous manner by means of electronic evaluation - Circuits or even in digital form by a run on a computer calculation software done.

Since the sensor signals can be subject to high-frequency interference, which in direct further processing and conversion into the control signals would lead to false or undesirable violent reactions in engine behavior, the sensor signals are generally smoothed by Gradientenbegrenzern and / or filters, preferably as a filter low-pass filter first or higher order.

When designing a filter for smoothing sensor signals, different, sometimes contradictory, ones must be used

Requirements are taken into account in order to ensure a quiet engine behavior on the one hand and to respond quickly enough to changes in the driver's request or the ambient conditions. to respond. Furthermore, it is to facilitate the effort for tuning the engine behavior in the test bench or. Driving test desirable that the filter is as easy to parameterize.

In summary, the following requirements are placed on such a smoothing filter:

- The filter should have low-pass behavior and thereby offer a good compromise between the attenuation of interference and a small delay of the actual sensor waveform.

- Under no circumstances should the filter overshoot.

- The filter should contain no dead time if possible.

- The filter should also be combinable with a gradient limitation.

- The filter should be easy to parameterize.

A known refinement of a low-pass filter which fulfills almost all of these requirements is described in DE-19753996-A1 as a method for damping jerking vibrations or load blows of an internal combustion engine. In this filter, the converted into an internal setpoint value, for example, a fuel mass to be injected, converted sensor signal of an accelerator pedal encoder with an ITi member filtered and the output of the IT - - filter on the

Input fed back. However, resulting from this filter arrangement PT ₂ filter behavior leads to overshoots can occur, ie. that the filtered setpoint is greater than the input setpoint. In order to avoid this overshoot, the calculation of the PT2 filter is reinitialized whenever the sign of the difference between the setpoint and the filtered setpoint changes. this leads to to discontinuities in the course of the filtered setpoint, which occur even when the setpoint applied to the filter input is continuous, for example, ramped. An unsteady filtered setpoint for the motor control following the desired value filtering means that unsteady actual values are also set, which can lead to inhomogeneous motor behavior and noise generation.

Object of the present invention is to provide a method of the type mentioned above taking into account all mentioned requirements and while avoiding the observed in the known design problems.

This object is achieved by a generic method with the steps of claim 1.

According to the sensor signal to be smoothed is filtered with a high-pass filter, then the time constant of the low-pass filter first order is determined with the current value of the output of the high-pass filter and then the sensor signal with the low-pass filter, which now works with the newly determined time constant, filtered.

By choosing a low-pass filter of the first order, almost all of the requirements are met. Thus, a first-order low-pass filter does not oscillate, it also has no dead time and can be combined with a gradient limiter without any problems. The use of the high-pass filtered sensor signal for calculating the time constant of the low-pass filter offers the further advantage that the entire filter arrangement responds to a jump-shaped sensor signal with an adjustable initial and an adjustable final behavior. In other words, both the initial increases tion and their temporal transition into the phase of the maximum slope as well as the temporal transition to the final slope by appropriate selection of the parameters of the high-pass filter and the calculation formula for the time constant of the low pass filter are set. In this case, the high-pass filter particularly determines the initial behavior of the step response since, depending on the choice of system-related deceleration, it reacts quickly or slowly to high change rates of the sensor signal. The final behavior is in turn determined by the choice of the time constant calculation formula, since a high time constant leads to a strongly delayed low-pass and thus a slow gradual transition into the final value of the step response, while a small time constant signifies a rapid transition to the final value - tet.

Furthermore, in the method according to the invention, the behavior of the filter arrangement can be adapted to the interference components in the sensor signal. The high-pass filter filters out the low-frequency components of the sensor signal, i. one receives a statement about the proportion of interference superimposed on the actual sensor signal. Depending on the selected relationship between the output signal of the high-pass filter and the time constant, the frequency-dependent course of the gain of the low-pass filter can now be determined, whereby the attenuation of the respective frequency components in the sensor signal is determined.

In a preferred embodiment of the method according to the invention, the current value of the output signal of the

High-pass filter linear in the time constant of the low-pass filter, wherein the time constant of the Tiefpas sfilters by multiplying the current value of the output signal of the Highpass filter is determined with a first parameter or by adding the current value of the output signal of the high-pass filter with a second parameter or by an overlay of the multiplication and the addition. Each of the possible linear links means that in the presence of higher frequencies and thus disturbances in the sensor signal, the time constant of the low-pass filter is increased, with the result that higher frequencies are attenuated more. In return, the time constant is reduced at low-frequency signal waveforms, so that the low-pass filter can follow the actual, less disturbed sensor signal quickly enough.

If either only one multiplication or only one addition is carried out, then the parameter set of the filter arrangement is only extended by an additional parameter. The parameter set otherwise contains the parameters of the high-pass filter and the gain factor of the low-pass filter of the first order.

With only one further parameter, therefore, the effort to adjust the filter arrangement during the experimental and

Setting phases of the internal combustion engine in conjunction with the

Control unit as low as possible.

The superimposition of the multiplication and addition offers an additional advantage, since it allows the setting of the start and end behavior of the step response of the filter arrangement almost independently of one another. Namely, with the first parameter, the multiplicative factor, the output signal of the high-pass filter is weighted, which results in a relative change in the time constant of the low-pass filter, depending on how large the output signal of the high-pass filter is. This results in an influence on the initial behavior of the step response, since a reduction of the first parameter Thus, the initial slope of the step response flattens and the initial behavior is delayed as a whole. The final behavior is not changed with the first parameter, since then the input signal of the filter arrangement, ie the jump, is at a constant value, which corresponds to a change speed of zero.

By contrast, the second parameter, the summand, has a greater effect on the final behavior of the step response since it affects the time constant of the low-pass filter absolutely and not only relatively. Regardless of the current size of the output signal of the high-pass filter, ie over the entire time course of the step response, the low-pass filter thus operates with greater or lesser delay, which is clearly visible in the final behavior.

The direct linking of the initial and final behavior of the jump response with one parameter in each case greatly simplifies the work of a test or driving test engineer who is to optimize the behavior of an internal combustion engine by parameterizing the electronics or the software of the associated control device.

In the linear embodiments of the method according to the invention, the absolute value of the output signal of the high-pass filter is preferably included in the calculation of the time constant, since only the magnitude of the output signal is of interest. Negative time constants can be excluded from the outset. In certain cases, the absolute value formation can, of course, only take place as the last step, following the calculation of the time constant, for example, when the output signal is only multiplied by a factor.

If one chooses as a high-pass filter. First-order filters, the high-pass filter with only a few parameters contributes to the scope of the parameter set. The parameter set and thus the effort for the design or. Adjustment of the filter arrangement can be further reduced by choosing a first-order low-pass filter with the unity gain. An amplification factor of one will always be preferred if the sensor signal is already present in the desired order of magnitude and in fact only smoothing should be performed. However, a gain factor not equal to one may be desirable if, for example, a conversion to another physical value range is still to be carried out or the resolution of the filtered sensor signal is to be improved.

The invention will be explained in more detail below with reference to an exemplary embodiment and the drawing, in which:

Fig. 1 is a occurring in the known filter arrangement

Ramp response;

FIG. 2 shows a block diagram of the filter arrangement according to the invention; FIG.

FIG. 3 shows the frequency response of the gain of a low-pass filter and a high-pass filter of the first order; FIG.

4 shows a graphical representation of a calculation function for the time constant of the low-pass filter; 5 shows different step responses with variation of a first parameter; 6 shows different step responses with variation of a second parameter;

Fig. 7 Simulation with real data.

FIG. 1 shows a ramp response of the known filter arrangement of feedback ITi filter, wherein the filter is always reinitialized as soon as the output value is equal to the input value of the filter. Shown is the waveform of the requested by a driver of a motor vehicle by pressing the accelerator pedal engine torque T _q _soll over time. The requested engine torque. In this case, T _q _soll is determined via an accelerator pedal sensor which measures the position of the accelerator pedal and fed to a motor control unit as an input signal, where it is converted or converted into the variable T _q _soll. Disturbances on the position signal are passed on undamped during the conversion. The waveform shown in Figure 1, however, shows an ideal, smooth ramp for T _q _soll, which is trouble-free. It would be desirable to have a filter response T _q _soll_filt which follows the ramp with slight delay. In the known design, however, clearly visible discontinuities occur, which are due to the reinitialization of the ITi filter. In FIG

1 is still differentiated between three different work areas: Area I means a reinsertion after thrust (?), Area II transmission system change (??) and only in area III works the known filter arrangement without interruption.

To avoid the discontinuities and to reduce the significant delays between the initialization points, the invention proposes a filter arrangement according to FIG

2, in which a sensor signal or an input signal x of a motor control unit with a high-pass filter HP filtered becomes. The output signal Xh _{P of} the high-pass filter HP is supplied to a function f (x _hp ). With the function f (x _hp ), the currently valid time constant T _T p of a first-order low-pass filter TPi is determined from x _h p. The sensor signal x is smoothed with the low-pass filter TPi to the signal xπ it.

FIG. 3 shows the amplification characteristics of a low-pass filter TPi and of a high-pass filter HPi, both filters being of the first order, in each case one filter gain (V _TP or v _H p) of one and a respective time constant (T _TP or T _HP ) of 1 sec, resulting in a respective pole (p _T p or P _HP ) at -1. The transfer functions of the two filters with the frequency variable p = j-ω thus have the following form:

V, TP _ __j__ and V HP ^■ P

^G HP = (1)

'TP ^■ p + \ / 7 + 1 ^■ p + lp + l

The amplification curves in FIG. 3 once again illustrate that the low-pass filter TP ₁ transmits signal components with low frequencies up to near the corner frequency of <B _TP = 1 / T _TP undiminished and then suppresses them more and more with increasing frequency, while the high-pass filter HPi with lower signal components Frequencies is suppressed, its gain then approaches more and more of the overall gain V _HP with increasing frequency, and it approximately from the corner frequency

the higher frequencies are filtered through.

From the output signal Xh _{P of} the high-pass filter HP, the time constant T _{TP of} the low-pass filter TP _{x is} determined according to the invention. FIG. 4 shows the diagram of the functional course of a ner preferred embodiment of the invention, wherein between x _hp and T _{rμ the} following relationship applies: f (x _hp ) = T _rp = Ax _hp + B. (2)

As can be seen in FIG. 4, the time constant T _{TP increases} with increasing x _hp , d. H . with increasing rate of change of the input signal x, too, so that the low-pass filter can attenuate higher-frequency interference better. At lower rates of change, the time constant T _τ p is again reduced, so that with a smaller amount of interference, the actual signal course of the input signal x can be better followed.

How changes in the parameters A and B affect the overall behavior of the filter arrangement of high-pass filter HPi, low-pass filter TPi and function f (x _hp ) in accordance with equation (2) can correspond to the step responses of the filter arrangement depicted in FIGS. 5 and 6 be removed. In FIG. 5, the parameter A is enlarged along the direction of the arrow shown. This leads to an increase in the influence of Xh _p on the time constant T _TP (see equation (3)), so that the step response is slowed down in the initial range, ie in the range of the high rate of change of the input signal x. The final behavior remains unaffected by parameter A.

FIG. 6 shows the influence of parameter B on the final behavior. As the parameter B increases, the time constant T ^ p is increased to be absolute, slowing down the low-pass filter TP _x over the entire operating range, irrespective of the actual rate of change of the input signal x, which is easy in the initial slope and particularly strong, becomes visible in the final behavior of the step response.

In the example according to FIGS. 5 and 6, the parameters A and B are changed exclusively, while the parameters T _HP and V _H μ of the high-pass filter HP i are kept constant. If further adjustments of the transmission behavior of the entire filter arrangement are required, these can be made via the choice of T _HP and VHP, wherein in particular the change of THP makes sense. A change of V _HP indirectly corresponds to a change of the parameter A, since Xhp is weighted with it.

The transfer functions of the high-pass filter HP ₁ and the low-pass filter TP ₁ according to the equations (1) map the transmission behavior of the two filters in the analog frequency range. As already explained, it is possible to implement the filter arrangement according to the invention in an engine control unit either analogously, by electronic circuits, or discretely, by calculation functions of a computing unit. For the discrete calculation of the filters HP ₁ and TP ₁ , the following difference equations apply, the gains V _HP and V _{TP being} respectively set to one, the time constants T _HP and T _T p remaining variable, and the parameter v being the respective time resp , Calculation cycle called:

HPi: x _hp (v) = r, ■ ( _xlψ (V-V) -x (v-V) + x (v)) (3)

TPi: x _filt (v) = z ₀ - X _j111 (VT) + (IZ ₀ ) ■ x (v) ^' (4).

The parameter r _α represents the discrete counterpart to the analog

Time constant T _H p and z _{0 is} the discrete counterpart to the analog time constant T _V p, whereby the known, in the discretization apply analogous transfer functions relationships to be considered.

Figure 7 shows a comparison of the known method with the particular embodiment of the method according to the invention. Again, the example of the driver request in a motor vehicle with internal combustion engine was chosen. The signal of the accelerator pedal converted into the required engine torque T _q _soll is filtered both with the known ITi filter arrangement, from which the signal T _q _soll_filt_SdT arises, and with the arrangement of HP _× filter, function f (x _hp ) Equation (2) and TPχ filter whose output signal is the signal T _q _soll_filt. The filtered signals are both also still guided over a Gradientenbegrenzer, which can be seen in particular in the period between 51.6 and 52 sec. FIG. 7 illustrates the improvements which are achieved in the filtered signal profile T _q _soll_filt by the method according to the invention, in particular the avoidance of the discontinuities and the reduction of the average delay, which becomes particularly clear in the time range between 51 and 51.5 sec.

Claims

claims

1. A method for smoothing a sensor signal, in particular an input signal of a control unit of an internal combustion engine, with a low-pass filter of the first order, characterized by the steps:

Filtering the sensor signal with a high-pass filter,

Determination of the time constant of the low-pass filter as a function of the current value of the output signal of the high-pass filter,

- Filtering the sensor signal with the low-pass filter.

2. The method according to claim 1, characterized in that the output signal of the high-pass filter is received linearly in the time constant.

3. The method according to any one of the preceding claims, characterized in that the time constant is determined by multiplying the current value of the output signal with a first parameter.

4. The method according to any one of the preceding claims, character- ized in that the time constant is determined by adding the current value of the output signal to a second parameter.

5. The method according to any one of the preceding claims, characterized in that the absolute value of the output signal is received in the time constant.

6. The method according to any one of the preceding claims, characterized in that the high-pass filter is a first-order filter. Method according to one of the preceding claims, characterized in that the low-pass filter has the gain one.