JP2001099678A - Electronic control apparatus - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an electronic control apparatus whose controllability is enhanced by shortening the time up to a conversion operation when floating point-type data is smoothed and computed. SOLUTION: An engine control ECU 10 fetches the detection value (the actual value) of a sensor 2 which is installed at an engine 1, and it executes the control of the injection of a fuel and the control of the purge of a fuel gas. A floating point computing processor (FPU) 15 inside the ECU 10 executes a floating point-type computing operation. When the ECU 10 smoothes and computes floating point-type data on the detection value of the sensor, it finds the inclination of a smoothed and computed result. When the absolute value of the inclination becomes a prescribed value or lower, a smoothed value is made to agree forcibly with the actual value.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an electronic control unit for calculating floating point data.

[0002]

2. Description of the Related Art In an electronic control unit (ECU) applied to the control of an engine for a vehicle, a smoothing operation is performed as a method for avoiding a sudden change in raw value. That is, for example, the actual value (raw value) of the input that constantly changes according to the operating state, such as the intake air amount and the throttle opening, fluctuates greatly. If this value is used for control as it is, the control system becomes unstable, Poor controllability. Therefore, for example, a smoothing operation is performed using the following equation to improve controllability. In this case, the average value = (previous average value-raw value) x average rate + raw value In this case, the difference between the average value and the raw value is multiplied by the average rate, and the product of the product and the raw value is calculated. The average value was determined by the sum. However, since the existing ECU treats the smoothed value by the integer type (fixed point type) operation, if the value of “(previous smoothed value−raw value) × smoothing rate” is smaller than the decimal point, The average value converged to the raw value due to the drop.

On the other hand, in recent years, it has been studied to introduce a floating point arithmetic processor (FPU: Floating-Point Unit) and to execute engine control using floating point type data. Floating-point data is composed of a sign part, an exponent part, and a mantissa part. For example, in the case of floating-point data consisting of four bytes, by having a 23-bit mantissa part, seven-digit precision (1/2 ^) is always obtained. The calculation is performed with 23 (the precision of 0.0000001), and very fine precision can be obtained compared to the conventional integer type data.

For example, when calculating the throttle opening in the detection range of 125 ° with an accuracy equivalent to 14 bits, the resolution (LSB) of the actual integer type is (125 deg / 16384 ≒ 0.00762).
9), whereas floating point type data is 76290 times (0.007629 / 0.0000001 ≒ 76290) compared to integer type
Will have the accuracy of

[0005]

In the case where an averaging operation is performed using floating-point data, the precision of the operation result is improved, and the number of lost digits in the conventional averaging operation is reduced. However, there is a problem that the smoothed value becomes too gentle as the smoothed value approaches the raw value, and it takes time to converge the smoothed value. Therefore, it is conceivable that the process of starting after the smoothed value converges is delayed, or the process that continues until the converges converges is unnecessarily prolonged, and controllability may deteriorate.

In other words, floating-point data has the property that a bit (such as an integer LSB) changes depending on its value. The cracks become thinner as the floating point type data approaches 0, and the digit loss at the time of calculation is small.
On the other hand, as the absolute value of the data increases, the roughness of the data also increases, and the degree of loss of digits during the calculation increases. In conventional integer arithmetic, convergence is achieved quickly by truncating fractions smaller than LSB, but significant digit loss occurs. Since the floating-point operation does not have the concept of LSB, the number of cancellations is small, but the operation does not easily converge to the operable precision.

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and has as its object to provide an electronic control unit capable of shortening the time required for convergence in smoothing operation of floating-point data. And

[0008]

In the electronic control device according to the present invention, when performing a smoothing operation on raw value floating-point data, a slope of a smoothing calculation result is obtained, and an absolute value of the slope is determined by a predetermined value. If the value is less than or equal to the value, the smoothed value is forced to match the raw value. Therefore, the problem that the convergence of the smoothed value is delayed and the controllability deteriorates due to the delay is solved. That is,
It is possible to reduce the time required for convergence during the smoothing operation of floating-point data and improve controllability.

In this specification, the term "raw value" refers to a value before smoothing, such as an actual detection value of various sensors or a control command value for various controls, and the present electronic control unit applies the raw value to the raw value. Performs an averaging operation and obtains an averaging value.

According to the second aspect of the present invention, the predetermined value for determining the slope of the smoothing operation result is set according to the required accuracy of the raw value. For example, in the case of a raw value for which the required accuracy is relatively low, if the predetermined value is increased, the accuracy decreases, but a smoothing operation with fast convergence becomes possible. In the case of a raw value requiring a relatively high accuracy, if the predetermined value is reduced, the convergence is slightly delayed, but the accuracy is improved accordingly.

According to the third aspect of the present invention, the predetermined value for determining the inclination of the result of the smoothing operation is set in accordance with the amount of change in the raw value. For example, when the change amount of the raw value is large, the predetermined value is made larger than when the change amount is small. This makes it possible to equalize the convergence time of the average value even if the amount of change in the raw value changes.

In this way, by appropriately changing the predetermined value for judging the slope of the smoothing operation result, it is possible to arbitrarily manage and adjust the accuracy of the smoothing value and the convergence time. it can.

[0013]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram schematically showing an engine control system. In the present system, the on-vehicle engine 1 is configured as, for example, a gasoline injection type multi-cylinder internal combustion engine.

The engine control ECU 10 includes a microcomputer (hereinafter, referred to as a microcomputer) 11, which includes a CPU 12, a RAM 13, a ROM 14, an FPU
(Floating point arithmetic processor) 15 and I / O (input / output device) 16. Here, the FPU 15 performs a calculation in a floating-point format, and the CPU 12 performs a calculation in a format other than the floating-point format.

Various information indicating the operating state of the engine is input to the ECU 10 from a sensor 2 provided in the engine 1. The sensor 2 includes, for example, an air flow meter for detecting an intake air amount, a water temperature sensor for detecting a cooling water temperature, a throttle opening sensor for detecting a throttle opening, and the like. Then, the ECU 10 controls fuel injection by an injector (not shown), purge control of fuel gas, and the like based on the input sensor signal.

The actual value detected by the sensor 2 (hereinafter referred to as the actual value) includes noise and the like and cannot be used for control as it is. That is, the actual value fluctuates greatly, and if this value is used as it is for control, there is a risk of malfunction. Therefore, the ECU 10 performs a smoothing operation on the input actual value to remove noise, and then uses the smoothed value for control.

The floating-point data calculated by the FPU 15 is formed, for example, in accordance with the IEEE754 standard. As shown in FIG. 2, the data consists of a 1-bit sign, an 8-bit exponent, and a 23-bit mantissa. And In the case of 4-byte floating point data in which the mantissa part is composed of 23 bits, it has a resolution of 7 digits (resolution of 0.0000001).

Next, the operation of the present embodiment will be described. FIG. 3 is a flowchart showing a time interruption process for taking in a sensor detection value. This process is executed by the CPU 12 in the engine control ECU 10 at a cycle of, for example, 8 ms. In the process of FIG. 3, the sensor detection value (actual value a)
Is calculated at predetermined intervals, and each time, either the actual value a or the average value b is output as a control value. In the course of such processing, the FPU 15 performs a floating-point operation on the real value a and the smoothed value b.

In FIG. 3, first, at step 110,
The actual value a is read from the sensor 2, and in the subsequent step 120, the actual value a is smoothed. That is, the averaging value b is calculated as the averaging value of this time = (the previous averaging value−the actual value) × the averaging rate + the actual value. At this time, the values of a and b are calculated by the FPU 15 in a floating-point type at high speed.

Thereafter, in step 130, the a, b
Is used to determine the slope c of the result of the smoothing operation. That is, the slope c is calculated as the slope of the smoothed calculation result = the current smoothed value−the previous smoothed value.

Thereafter, in step 140, it is determined whether or not the absolute value of the gradient c is equal to or smaller than a predetermined value kc. If | c |> kc, the average value b is output as the control value in step 150, and then the present routine is terminated. If | c | ≦ kc, the actual value a
Is output as a control value, and then this routine is temporarily terminated.

FIG. 4 is a time chart showing changes in the smoothed value. In FIG. 4, the actual value is a1 as shown in FIG.
From a to a2. At this time, in (b) exemplifying the conventional integer type operation, at time t1 when the smoothed value changes as shown in the figure and the update amount of the smoothed value becomes less than the LSB,
The smoothing operation is stopped, and the actual value a2 is set as the control value. That is, at time t1, the smoothed value converges to the actual value a2. Also,
Illustrates a conventional floating point operation (FPU operation) (c)
Then, the smoothing operation is continued even after the time t1 at which the smoothing operation is stopped in the above (b). That is, comparing the above (b) and (c), in the integer type operation of (b),
It converges quickly by truncating fractions less than LSB,
Digit drop is large. In the FPU calculation of (c), the LSB
Since there is no concept, the smoothing operation is performed up to the operable accuracy. For this reason, the number of digits lost is small, but it does not easily converge.

On the other hand, the floating point operation (FP
In (d) exemplifying U operation), the slope c of the smoothing value is obtained, and when the absolute value of the slope c becomes smaller than the predetermined value kc, the smoothing operation is forcibly stopped and the actual value a2 is changed to the control value. And Therefore, the same time t as the smoothing operation in the integer type operation
At 1, the actual value converges, and the convergence time of the smoothed value is shortened as compared with the above (c).

According to the embodiment described above, when smoothing floating-point data of real values (raw values),
The problem that the convergence of the averaging value is delayed and the controllability deteriorates due to the delay is eliminated. In other words, even if the control specification built by the conventional operation using the integer type is directly applied to the floating point, the convergence time is not prolonged compared to the case of the integer type operation, and the operation is started after the smoothed value converges. It is possible to avoid such a problem that such processing is delayed or processing that continues until convergence is unnecessarily prolonged, thereby deteriorating controllability. Therefore, controllability equivalent to smoothing operation with integer data can be secured.

The present invention can be embodied in the following forms other than the above. A predetermined value kc for determining the slope c of the smoothed value is set according to the required accuracy of the sensor detection value (actual value). For example, in the case of a sensor detection value with a relatively low required accuracy, such as a water temperature, the predetermined value kc is increased to give priority to convergence over the accuracy of the smoothed value, and a sensor detection value with a relatively high required accuracy, such as an intake air amount, is detected. In the case of a value, the predetermined value kc is made smaller to give higher priority to the accuracy improvement than the convergence of the smoothed value.

Further, a predetermined value kc for determining the slope c of the smoothed value is set according to the amount of change in the sensor detection value (actual value). For example, when the amount of change in the sensor detection value is large, the predetermined value kc is made larger than when the amount of change is small. As a result, even if the amount of change in the sensor detection value changes, the convergence time of the smoothed value can be made uniform.

In this way, by appropriately changing the predetermined value kc according to the required accuracy and the amount of change in the sensor detection value, the accuracy of the smoothed value and the convergence time can be arbitrarily managed and adjusted.

In the above-described embodiment, the case where the sensor detection value is smoothed is specified. Alternatively, the control command value for various actuators (driver circuits) may be smoothed. In this case, for example, in the case of a control command value having a relatively low required accuracy, such as a purge control of fuel evaporative gas, the predetermined value kc for determining the slope c of the smoothed value is set to be larger than the accuracy of the smoothed value. In the case of a control command value with relatively high required accuracy, such as fuel injection control or electronic throttle control, it is preferable to reduce the predetermined value kc and give priority to improving accuracy over convergence of the smoothed value. Further, the predetermined value kc may be set variably according to the amount of change in the control command value.

[Brief description of the drawings]

FIG. 1 is a block diagram schematically showing an engine control system according to an embodiment of the present invention.

FIG. 2 is a diagram showing a configuration of floating-point type data.

FIG. 3 is a flowchart showing a time interruption routine by a CPU.

FIG. 4 is a time chart showing transition of an average value.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Engine, 10 ... Engine control ECU (electronic control device), 11 ... Microcomputer, 12 ... CPU, 15 ... FPU
(Floating point processor).

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 2F024 AA02 AF01 3G084 BA13 BA15 BA27 DA06 EA08 EA11 EB02 EB07 EB25 EC04 FA07 FA10 FA20 5B022 AA00 BA01 EA07 FA01 FA07

Claims (3)

[Claims] An electronic control unit for calculating floating-point data, wherein, when smoothing raw-valued floating-point data, a slope of a smoothing calculation result is obtained, and an absolute value of the slope is set to a predetermined value or less. An electronic control unit forcing an annealed value to coincide with a raw value. 2. The method according to claim 1, wherein the predetermined value for determining the slope of the result of the smoothing operation is set in accordance with a required accuracy of the raw value.
An electronic control unit according to claim 1. 3. The electronic control device according to claim 1, wherein the predetermined value for determining the inclination of the result of the smoothing operation is set according to a change amount of the raw value.