DE102014216312A1 - Method for sensorless temperature measurement of a DC motor - Google Patents

Abstract

Proposed is a method for determining the temperature of the electric coil of a DC motor, wherein the temperature is determined sensorless with the following method steps: - Determining the time constant τ = L / RL of the coil, where L is the inductance and RL is the ohmic resistance of the coil, - Determination of Coil temperature using the time constant τ and the temperature characteristic of the ohmic resistance RL of the coil.

Description

For most applications in mechanical engineering, certain system areas must be specifically cooled with different liquid or gaseous coolants. In vehicles, a clever cooling process can enable a huge reduction in fuel consumption and CO ₂ emissions. For this purpose, intelligent cooling water circulation between several parts, such. B. Cylinder crankcase, engine radiator and exhaust gas turbocharger, take place. Until a few years ago, the cooling water was pumped through the various parts without significant controlled metering, with some parts being under-cooled and others too much. In addition, it is advantageous to heat some areas, such as the engine oil at engine start quickly to temperatures against 125 ° C, ie to supply the heat from the other areas of the engine oil.

For less than a decade, internal combustion engines have been equipped with intelligent thermal management systems. A sensor-controlled rotary valve allows a quick achievement of the ideal temperature window for the engine and transmission. A spherical valve having multiple passages is rotated by a DC motor and placed in predetermined positions. Almost all cooling water flows through the spherical valve. Each of the vehicle parts to be cooled is assigned to one of the passages of the spherical valve. The intelligent rotation and positioning of one of the passages of the spherical valve, the associated area is flowed through by the coolant and consequently selectively cooled.

The above-mentioned DC motor is usually operated by means of PWM signals from a control unit of the vehicle, which is located outside of the rotary valve module. Heavy load on the electric motor and / or its high ambient temperature can lead to its unexpected and unwanted defect and failure. Permanent temperature monitoring of the inside of the engine makes perfect engine operation indispensable. By mounting a temperature sensor on the motor housing, it is possible to detect the external motor temperature, which differs from the internal temperature of the motor. However, housing the temperature sensor in the engine could allow a more accurate temperature determination of the sensitive inner parts. Both methods require the use of not so inexpensive temperature sensors. In addition, a problem is the placement of the temperature sensor in the small seductive space of the electronic control unit of the rotary valve. It should be noted that a location of the temperature sensor in the already small engine is very expensive, which leads to the increased price of the engine.

The present invention has for its object to eliminate the above-described lack of a favorable method for measuring the temperature of the DC motor. According to the invention, the temperature dependence of the electrical parameters of the coils in the motor is used. By means of the method according to the invention, the motor temperature can be determined by determining the resistance and optionally the inductance of the coil. The winding wire of the coils is typically copper whose resistivity increases with temperature and has a fairly good linear temperature dependence.

The invention will be explained below in embodiments and with reference to the following drawings. Showing:

1 the temperature dependence of the resistivity of copper in the automotive-relevant temperature range;

2 an equivalent circuit diagram of a motor string;

3 a rectangular periodic PWM signal: a) the voltage U (t), b) the current I (t) as a function of the time t;

4 Courses of the function (1 - e ^x ) and their first four series developments;

5 Determination of the time constant τ with the aid of two current measuring values;

6 Circuit with a DC motor and a measuring resistor (shunt);

7a ) experimental current flow as a function of time, b) linear course of the time constant τ as a function of time t;

8th Flowchart for the method and the algorithm for determining the time constant and the motor temperature.

1 shows the temperature dependence of the resistivity of copper in the automotive-relevant temperature range between -40 ° C and 150 ° C, where a slope of 0.0068 × 10 ^-8 Ωm / K is observed ( RA Matula: Electrical Resistivity of Copper, Gold, Palladium and Silver; J. Phys. Chem. Ref. Data; Vol. 8, no. 4, pages 1147-1298, 1979 ). In addition, the coils in most inexpensive DC motors contain magnetizable cores whose magnetic permeability is also temperature-dependent. In general, the permeability increases with the temperature up to a maximum, then drops to the value 1 at the Curie temperature. This temperature dependence is reflected in the inductance of the coils, the inductance of the permeability proportional (hereinafter, for the sake of simplicity, is referred to as the coil in the singular). If the coil does not contain a magnetizable core ("ironless"), its inductance should have no temperature dependence. In that case only the temperature dependence of the winding wire is considered. If the temperature dependence of the ohmic resistance and the inductance of the coil is known, the internal temperature (in the motor) can be determined by determining the resistance and the inductance of the coil.

Each engine train can be simplified by the in 2 illustrated circuit diagram, which applies in principle for all permanent magnet synchronous machines, to be replaced. Where U is the applied DC voltage, L is the coil inductance and R _{L is} the ohmic resistance of the coil. Due to the motor rotation, the voltage U _{ind is} induced in the motor string by the permanent magnets. The induced in the motor string voltage U _ind (or armature voltage) is proportional to the engine speed and remains constant as long as the engine speed and the engine load remain stable or constant. The following equation then applies to the build-up and -down of the circuit:

wobei U₀ für den Hochpegelwert des Pulses ist (s. 3a). Der Koeffizient bzw. die Zeitkonstante τ = L/R_L in Gl. 2 ist das Maß für die Trägheit des Stromverlaufs I(t), was sich in der Steilheit und Krümmung der Kurve in jeder Halbperiode widerspiegelt. Mit dem erfindungsgemäßen Verfahren wird daraus die Zeitkonstante R_L/L ermittelt.Since, as discussed above, the rotary valve is driven by periodic PWM signals, the voltage U has a rectangular shape (s. 3a ): At each sharp edge of the PWM signal, the derivative of the current I may become non-zero. This means that the assembly and disassembly of the current I get some inertia, which in Eq. 2, the solution of the differential equation Eq. 1 and off 3b to recognize is:

where U _{0 is} the high level value of the pulse (s. 3a ). The coefficient or the time constant τ = L / R _L in Eq. 2 is the measure of the inertia of the current waveform I (t), which is reflected in the slope and curvature of the curve in each half cycle. With the method according to the invention, the time constant R _L / L is determined from this.

wobei

ist.If one measures the current value I ₁ = I (t ₁ ) and at the time t ₁ and its value in Eq. 2, you should be able to practically determine the time constant τ. The problem is the other unknown values of U _ind and R _L. To avoid this, the above-mentioned measurement is carried out twice at the times t ₁ and t ₂ . One then obtains equations 3 and 4:

in which

is.

By dividing the two sides of Eq. 1 by the Gl. 2, the unknown factor C is shortened and Eq. 5 is resulted. The measured values t ₁ , t ₂ , I ₁ and I _{2 are} known, while only the time constant τ is unknown.

A solution of Eqs. 5 is possible by the Taylor series expansion of the exponential function:

4 shows the curves of the function (1 × e ^x ) and their first four series developments as a function of x with the time constant τ = 0.6 ms. For the following description, the series expansion of the 2nd degree has been used, whereby a good coverage of (1-e ^x ) and the series expansion at x <0.2 can be seen (see the hatched area in FIG 4 ). In that case, the solution of Eq. 5, for the time constant τ, the following:

To calculate the time constant τ using Eq. 7, at least two measurements of I (t) must be made at t ₁ and t ₂ , as in 5 shown to be determined. The measurement points, as described above and discussed later, should preferably be less than 0.2τ _m , where τ _{m is} the lower limit of the expected time constant.

For applications with high-performance hardware and high computational power, it is possible to use even higher levels of series expansion or even perform numerical calculations, which promises greater accuracy for the solution, but the calculation and implementation costs are considerably higher. In the following it will be shown that the 2nd order series expansion provides good accuracy by the method of the present invention.

First of all, however, we will deal with the method of current measurement, since the determination of the current values (see, for example, equation 7) is necessary. An accurate and simple current measurement is usually realized by means of a shunt resistor. In this case, a shunt, which has a very low resistance (mΩ) and very low temperature dependence, connected in series with the motor.

The current measurement can certainly also be carried out in another way (eg by means of non-contact current measurement as in DE 10 2012 009 243 B3 and in the references there). The main thing is that the current flowing through the motor is measured.

6 shows a simple circuit with a DC motor, a shunt (R _sh ) and a voltage source U ₀ with rectangular AC voltage. The voltage differences at the two ends of the shunt and at the two motor terminals are simultaneously measured by means of a detection unit. Since the resistance value of the shunt is known, the motor current I can be determined by dividing the shunt voltage U _sh by the resistance value R _sh . As a detection unit, a so-called instrument amplifier is usually used, which includes three operational amplifiers and has a high sensitivity and temperature stability. Sensitivity is very important because R _{sh is} too small, and therefore the Shunt voltage U _{sh is} in the mV range. The shunt voltage U _sh then has the same exponential form as the current I (t) in 3b , By the detection unit, the shunt voltage U _sh is amplified by several orders of magnitude (with the amplification factor A _v in 6 ) and forwarded to the evaluation unit. The evaluation unit may for example be a microprocessor which receives the signal via its analog-to-digital converter (A / D). By means of the algorithm of the present invention, the magnitude of the curvature of this exponential function, ie the time constant, shall be determined. This algorithm (see below) can be B. be implemented as software in the microprocessor.

If one multiplies in Eq. 7 denominator and counter with R _sh , then Eq. 8, in which instead of I ₁ and I ₂ , the voltage measurements U _sh1 = I ₁ R _sh , and U _sh2 = I ₂ R _sh are used:

In 7a the experimental U _sh (t) is plotted as a function of time t. In order to measure the above-mentioned curvature or to determine the time constant τ, the signal is measured or sampled at several (at least two) points in time. These sampling points are in 7a marked from a to f. The number of sampling points and their time differences or intervals .DELTA.t are set due to the requested measurement accuracy and resolution. By using the measured values of U _sh (t) and t in Eq. 8, the values of the calculated time constant τ can be calculated from sampling point b up to and including f (between times t _a and t _b up to and including t _e and t _f ). In 7b the linear course of τ is plotted as a function of time. It can be seen that no constant value is determined from this. The reason, as mentioned above and also out 6 it can be seen that the function (1 - e ^x ) does not completely resemble their series developments. In the series developments of all degrees by very small time values, the calculated time constant approaches its correct value (best at t = 0). Due to the (linear) interpolation of the measured values and their (linear) extrapolation, the intersection point on the Y-axis can then be determined, which is then to deliver the sought time constant ( 7b ).

From this analysis results for the example in 7 a time constant equal to 0.5931 ms, which deviates only 1.15% from the actual value (0.6 ms), which is acceptable for many applications. If higher accuracy is required, the number of sample points should be increased and the first sample point (ie sample point a) take place earlier (ie, t _a → 0). This is best in the inset of 7b seen. The inset shows, for comparison, two curves for the time constant: The dashed line is the curve with t _a = 20 μs (s. 7a ), while the solid curve represents the curve with t _a = 0.1 μs. A very slight curvature appears in the line when t _a → 0, so the time constant with a value of 0.6002 tends to be closer to the actual value of 0.6.

In 7a only the measuring points at t ≤ 1.2 ms have been taken into account for the analysis, ie at sampling points less than about t _max = 0.2τ, with τ = 0.6 ms, as discussed above (here t _max = t _f ). If, for example, values between 0.4 ms and 0.8 ms are to be expected for an application due to the temperature changes for the time constant, the analysis must best be carried out for t <t _max = 0.08 ms (ie 0.2 × 0, 4 ms = 0.08 ms). If, for the analysis described above, the data were used at t <t _max = 0.3 ms (ie half as much as the time constant to be determined), a value equal to 0.5915 ms would have been obtained for the time constant, ie with a deviation of 1.4%, greater than 1.09% (see above), but still with a not so big difference. This means that even higher t _max values can be used for an application with higher tolerances. In that case, the request to the evaluation unit becomes smaller since the resolution of the time measurement may be smaller. In principle, it is empirically possible to determine a zero setting (offset) for each application, which is then taken into account in the calculation of the time constant in order to further reduce the abovementioned deviation.

8th Figure 4 shows the flowchart for the method and algorithm for determining the time constant and the engine temperature of the present invention.

mit

It may possibly be difficult for a slow acquisition unit to perform its sample points, especially the first sample point shortly after the signal rise. In that case, the series expansion should not be computed by x = 0 (as in Equation 9) but by a point x ₀ that is not close to x = 0. As a result, Series Development looks like this:

With

In that case, the solution equation for the time constant τ is of a second degree and more complicated than the simple Eq. 7 or 8, consequently higher computing power is required for the evaluation unit.

By means of the device of the present invention, it is possible to determine the internal temperature of a DC electric motor driven by a PWM signal without using a common temperature sensor. The motor temperature is determined by the determination of the ratio of the inductance L and the resistance R _L , ie the time constant of the motor coil. In other words, the two mentioned electrical engine parameters with known temperature characteristic curves are used as the temperature sensor. The method is not applicable when R _L and L have the same or similar temperature characteristics, because, in that case, the temperature dependence of their ratio is very small and the temperature measurement in this way is subject to low resolution. Most coil wires are made of copper whose resistivity, as discussed above, increases with temperature. The temperature dependence of L is dependent on the permeability of the magnetizable coil core and can be defined and tailored by means of targeted change of the alloy constituents and their nature. A falling with the temperature inductance can lead to a higher temperature-dependent time constant than the temperature dependence of the copper itself or the usual temperature filler. Such alloys, for. As iron, silicone and aluminum, are available on the market, have with the temperature rise between the room temperature and 150 ° C, a reduction in the permeability of about 7%. In this case, in view of the increase of the coil resistance of approx. 50% (see above), a reduction of approx. 57% is to be expected for L / R _L , ie higher measuring resolution. If the coil is ironless, ie with non-magnetizable core, no temperature dependence is expected for L.

The variance of the temperature characteristics of R _L and L for a particular engine design must not exceed a predefined limit. In principle, instead of using independent and separately measured temperature characteristic curves of R _L and L, it is conceivable that the temperature dependence of L / R _L , ie the time constant, can be determined experimentally and used as a characteristic curve.

There are, as a rule, as indicated at the beginning, not one but several coils in the DC motor available. Not all coils are connected at the same time. During engine operation, one or several coils are separated by the brush contacts and one or the other is switched through. Due to these disconnections and connections, certain interference signals can occur in the U _sh cycle. Due to the fast motor rotation, however, a relatively low-noise U _Sh signal can be obtained.

The monitoring and limitation of the motor current is one of the most important requirements in the industrial v. a. automotive applications. Using the present method, both the motor current and the motor temperature can be determined. In this temperature measurement, there is no need for the use of temperature sensors whose mounting or placement on or in the engine are very expensive.

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited patent literature

• DE 102012009243 B3 [0024]

Cited non-patent literature

• RA Matula: Electrical Resistivity of Copper, Gold, Palladium and Silver; J. Phys. Chem. Ref. Data; Vol. 8, no. 4, pages 1147-1298, 1979 [0014]

Claims (1)

Method for determining the temperature of the electrical coil of a DC motor, characterized in that the temperature is determined sensorless with the following method steps: - Determining the time constant τ = L / R _{L of} the coil, where L is the inductance and R _{L is} the ohmic resistance of the coil, - Determination of the coil temperature with the aid of the time constant τ and the temperature characteristic curve of the ohmic resistance R _{L of} the coil.

Images