JPH04325194A - Fully automatic washing machine - Google Patents

Abstract

PURPOSE:To determine and execute the water flow strength of a washing operation and the washing time by the neuro-control to which a weight coefficient and a threshold are set by learning in advance. CONSTITUTION:Cloth quantity data D1, cloth quality data D2, dirt quantity data D3, dirt quality data D4, detergent quantity data D5, detergent quality data D6, and water temperature data D7 are inputted to a neural net 45, and by obtaining water flow data O1, washing time data O2, and revolution speed data O3 of a stirring body by its operation, the washing operation is executed.

Description

[Detailed description of the invention]

0001 [Purpose of the invention]

0002

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a fully automatic washing machine which executes a washing operation using neurocontrol.

0003

[Prior Art] Conventional general fully automatic washing machines are equipped with multiple washing courses such as a standard course, a stubborn dirt course, a soaking course, and a hand-washing course, each of which has a different washing time and water flow strength. The washing courses are stored in a microcomputer installed in the washing machine, and a desired washing course is selected and executed using a course switching key.

[0004]Furthermore, a fully automatic washing machine using fuzzy control that has recently become popular is one that detects the fabric conditions of the laundry, such as the amount of fabric, fabric quality, etc., and determines and controls the washing time and water flow intensity. It is.

[0005]

[Problems to be Solved by the Invention] In conventional general fully automatic washing machines, the condition of the laundry, that is, the amount and quality of the fabric, and the amount and quality of dirt, which are soiling conditions, vary greatly depending on household circumstances. Therefore, it is impossible to deal with these problems using a predetermined washing course, and it is not possible to obtain the optimum washing effect.

[0006] In addition, in the conventional fully automatic washing machine using fuzzy control, the factors that determine the washing time and water flow intensity are as follows:
We take into consideration the fabric conditions such as fabric amount and fabric quality, but in addition to that,
Since soiling conditions such as soiling amount and soiling quality are also considered, a more optimal washing effect cannot be expected by controlling only by the cloth conditions.

The present invention has been made in view of the above circumstances, and its object is to provide a fully automatic washing machine that can obtain optimal washing effects depending on the condition of laundry etc.

[0008] [Structure of the invention]

[0009]

[Means for Solving the Problems] The fully automatic washing machine of the present invention employs neurocontrol in which weighting coefficients and threshold values learned in advance are set based on input data from a sensor that detects the state of laundry, etc. The present invention is characterized by comprising a control means for determining and executing the water flow intensity and washing time of the washing operation.

0010

[Function] According to the fully automatic washing machine of the present invention, the water flow strength and washing time of the washing operation are determined and executed by neurocontrol according to the condition of the laundry, etc., so the condition of the laundry etc. varies. However, it is possible to cope with this, and the optimum washing effect can be expected.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS A first embodiment of the present invention will be described below with reference to FIGS. 1 to 22.

First, the overall configuration of a fully automatic washing machine will be described according to FIG. 2. An outer tank 2 is arranged inside the outer box 1. In the outer tank 2, a drive mechanism 4 mainly including a motor 3 is disposed at the outer lower part thereof, and a drain valve 5 and a drain hose 6 are also disposed. Further, an inner tank 7 is disposed inside the outer tank 2 and is rotated by a drive mechanism 4 during dewatering. Note that a large number of dehydration holes 8 are formed in the peripheral wall of the inner tank 7. Further, in the inner tank 7, an agitator 9 is arranged which is rotated by the drive mechanism 4 during washing and rinsing.

On the other hand, a top cover 10 is mounted on the outer box 1, and inside the rear of this top cover 10, there are
A water supply valve 11 that supplies water to the inner tank 7 (outer tank 2) is provided, and a water level sensor 12 that detects the water level in the outer tank 2 (inner tank 7) is provided. A microcomputer 13 as a control means that operates as described later is disposed inside the front part of the top cover 10.

Furthermore, an operation panel 14 as shown in FIG. 4 is provided at the front end of the top cover 10. This operation panel 14 is provided with a washing course display section 15, which displays "Standard", "Stubborn", "Soak", "Wash separately", "Speed", "Large" and "
Characters indicating each course of ``hand washing'' are displayed, and light emitting diodes 15a and 15b are displayed on the left side of each course.
, 15c, 15d, 15e, 15f and 15g are attached.

Further, the operation panel 14 is equipped with a course switching key 16, a start key (also used as a pause key) 17, a segment type display 18, and a neuron switch key 16. A light emitting diode 19 for sensor display is attached.

Now, a permeability sensor 20 is attached to the inner bottom of the aforementioned outer tank 2. This transmittance sensor 20
As shown in FIG. 5, the light-emitting element 22 made of, for example, a light-emitting diode and the light-receiving element 23 made of, for example, a phototransistor are disposed to face each other in a case 21, and the light-emitting element 22 in the case 21 The portions 21a and 21b where the light-receiving element 23 faces each other are made of a transparent material. In this case, the light reception signal of the light receiving element 23 is A/D converted and outputted as the transmittance signal S20 from the transmittance sensor 20.

Furthermore, a temperature sensor 24 is installed at the outer bottom of the outer tank 2.
is installed. As shown in FIG. 6, this temperature sensor 24 has a structure in which a thermistor 26 is disposed within an annular protrusion 25 formed in the outer tank 2, and the thermistor 26 is embedded and fixed with resin 27. Then, the detection signal of the thermistor 26 is A
/D conversion and output as a water temperature detection signal S24.

Now, the electrical configuration will be described according to FIG. An input port of the microcomputer 13 is connected to an output terminal of a key operation input unit 28 that outputs signals based on various key operations on the operation panel 14, and the other input ports are connected to a water level sensor 12 and a permeability sensor. 20 and the output terminal of the temperature sensor 24 are connected. Further, one input port of the microcomputer 13 is connected to the output terminal of a rotation speed sensor 28 for detecting the rotation speed of the motor 3, and the output terminal of a capacitance sensor 29 is connected to the other input port.

The output port of the microcomputer 13 is connected to the motor 3 and the water supply valve 11 via drive circuits 30 and 31.
The other output ports are connected to the light emitting diodes 15a to 15g and 19 via the drive circuit 32. One output port of the microcomputer 13 is connected to the display 18 via a drive circuit 33, and the other output port is connected to an electromagnet 35 via a drive circuit 34. In this case, the electromagnet 35 is
These are for the clutch of the drive mechanism 4 and for the drain valve 5.

In the above case, when the course switching key 16 shown in FIG. The light emission of the light emitting diodes 15a to 15g is shifted each time the course switching key 16 is pressed or while the course switching key 16 is pressed, and when the course switching key 16 is released, the light emission is stopped from shifting. Then, it is determined that the washing course corresponding to the light emitting diode that is emitting light at that time has been selected.

Now, the configuration of the capacitive sensor 29 will be described based on FIG. 7. The drive circuit 30 of the motor 3 includes an inverter 37 made up of a plurality of transistors 36 and the like, and a drive voltage Vac obtained by converting the power supply voltage of the AC power supply 38 to the motor 3 via the inverter 37.
is applied. The current Im flowing through the motor 3 is detected by a current transformer 39, and the detected current is converted into a voltage and applied to a non-inverting input terminal (+) of a comparator 40. Note that the inverting input terminal (-) of the comparator 40 is set to ground potential. Further, the drive voltage Vac is divided and applied to the non-inverting input terminal (+
), and the inverting input terminal (-) of the comparator 41 is set to ground potential. The output signals of the comparators 40 and 41 are sent to the exclusive OR circuit 4.
2, the output signal of the exclusive OR circuit 42 is given to the integrating circuit 43, and
The output signal of the integrating circuit 43 is applied to an input terminal of an A/D converting circuit 44.

[0022]The phase difference θm of the current Im flowing through the motor 3 with respect to the drive voltage Vac is designed to decrease as the output of the motor 3 increases.
The output is proportional to the amount of laundry (cloth amount). Therefore, by detecting the phase difference θm, it is possible to detect the amount of cloth.

That is, as shown in FIG. 8, comparators 40 and 41 generate pulses Vv according to current Im and drive voltage Vac.
and Vi, and in response, the exclusive OR circuit 42 outputs a phase difference pulse Vθ. Therefore,
The integrating circuit 43 calculates the average voltage Vθa of the phase difference pulse Vθ.
If detected, the A/D conversion circuit 44 outputs a digitalized cloth amount signal S29.

On the other hand, the microcomputer 13
It is equipped with a neural network 45 for neural control as shown in FIG. Although the neural net 45 is actually constructed in a simulated manner using software by the microcomputer 13, for convenience of explanation, it is shown here as being constructed using hardware.

That is, the neural network 45 has an input layer I consisting of seven units I1 to I7, an intermediate layer J consisting of five units J1 to J5, and three units K.
It consists of an output layer K consisting of K1 to K3, and an input layer I.
The units I1 to I7 of the intermediate layer J and the units J1 to J5 of the intermediate layer J are mutually coupled by links, and the units J1 to J5 of the intermediate layer J and the units K1 to K3 of the output layer K are connected to each other by links.
and are interconnected by links.

Next, FIGS. 9 to 2 will explain the operation of this embodiment.
This will be explained with reference to 2.

First, the laundry is put into the inner tub 7 together with the detergent, and the light emitting diode 15a is made to emit light by pressing the course switching key 16 shown in FIG. 4, thereby selecting the "standard" course as the washing course. select.

Hereinafter, referring to FIG. 9, when the start key 17 is pressed (time T0), the microcomputer 13, via the drive circuit 32, turns on a light emitting diode displaying the word "neurosensor". 19 to emit light. Moreover, when the start key 17 is pressed, the "washing" process starts, and the microcomputer 1
3, the water supply valve 11 is energized via the drive circuit 31, and water is supplied into the inner tank 7 (outer tank 2). Thereafter, when the water level in the inner tank 7 reaches a "small" level among the "high", "medium", "low" and "small" water levels, the microcomputer 13 controls the water supply valve 11 based on the signal from the water level sensor 12. (T1). At this time, the microcomputer 13 outputs the transmittance signal S2 from the transmittance sensor 20.
0 is read and stored in the RAM as detection data A1. In addition, the microcomputer 13 has a drive circuit 30.
The motor 3 is energized via.

In this case, the motor 3 is repeatedly energized for 0.6 seconds (forward rotation), de-energized for 1 second, energized for 0.6 seconds (reverse rotation), and de-energized for 1 second, thereby driving the agitator 9 to rotate. Thus, a detection water flow is generated. At this time, the forward and reverse rotational speeds of the agitator 9 are set to the "standard" speed of 72 rpm. Such a detected water flow is generated for about 15 to 20 seconds, during which time the microcomputer 13 reads the cloth amount signal S29 from the capacitive sensor 29 and obtains the cloth amount data D1.
The amount of cloth is determined based on this stored in the RAM. Thereafter, the microcomputer 13 cuts off the power to the motor 3 and energizes the water supply valve 11 again (time T2).

[0030] The water amount judgment by the microcomputer 13 is here based on three levels of ``small,''``medium,'' and ``large.'' Depending on each judgment, the water level is set to ``low,'' or ``large.''
Set to "medium" and "high" water levels. Thereafter, when the water level in the inner tank 7 reaches the set "low", "medium" or "high" water level, the microcomputer 13 cuts off the power to the water supply valve 11 and energizes the motor 3 (time T3). In this case, the motor 3 is repeatedly energized for 1 second (forward rotation), de-energized for 0.9 seconds, energized for 1 second (reverse rotation), and de-energized for 0.9 seconds, and the agitator 9 generates a standard water flow. be done. Also at this time, the forward and reverse rotational speeds of the agitator 9 are set to the "standard" 72 rpm. When this is carried out for a predetermined period of time, the motor 3 is cut off by the microcomputer 13 (time T4), and then,
It is stopped for a relatively short period of time (eg, 15 seconds). During this stop period (between times T4 and T5), the microcomputer 13 changes the light emitting diode 19 of the "neurosensor" from continuous lighting to blinking lighting so that the user does not misunderstand that the stoppage is due to a failure.

When the motor 3, that is, the agitator 9 is stopped, the movement of the washing liquid stops, air bubbles, etc. rise, and the vibrations of the outer tub 2 also disappear, so that the permeability of the washing liquid becomes stable. do. Therefore, the microcomputer 13 reads the transmittance signal S29 from the transmittance sensor 29 and stores it in the RAM as detection data A2. Then, as shown in FIG. 11, the microcomputer 13 compares the above-mentioned detection data A1 and the current detection data A2, and if the difference is small, determines that the injected detergent is a liquid detergent,
If the difference is large, it is determined that it is a powdered detergent, and the result is stored in the RAM as detergent quality data D6.

That is, liquid detergent does not make the washing liquid so cloudy, but powder detergent becomes more cloudy as it dissolves in the washing liquid. Thereafter, the motor 3, that is, the stirring body 9, is rotated with the standard water flow (time T5).

When the agitator 9 is rotated in the forward or reverse direction, the number of revolutions of the motor 3 changes depending on the type of laundry, that is, the quality of the fabric, each time the agitator 9 is activated. The microcomputer 13
The amount of change in the rotation speed of the rotation speed signal S2 of the rotation speed sensor 28 is
8, the fabric quality is determined based on the magnitude of the change, and is stored in the RAM as fabric quality data D2 such as "stiff", "standard", and "supple" (time T6).
). Also, at this time, the microcomputer 13 reads the temperature signal S24 from the temperature sensor 24, and stores the water temperature of the washing liquid in the RAM as water temperature data D7 based on this.

[0034] When the microcomputer 13 determines the fabric quality, the microcomputer 13 sets the water flow to ``strong'' or ``standard'' depending on the fabric quality of ``stiff,''``standard,'' or ``flexible,'' for example. Or set it to "weak" water flow. As a result, when the microcomputer 13 determines that the water flow is "weak", it energizes the motor 3 for 0.9 seconds (normal rotation), and energizes the motor 3 for 1.1 seconds.
The power is cut off for 0.9 seconds, the power is turned on for 0.9 seconds (reverse rotation), and the power is cut off for 1.1 seconds repeatedly. When it is determined that the water flow is "standard", it is as described above, and when it is determined that the water flow is "strong", the motor 3 is energized for 1.1 seconds (forward rotation), energized for 0.7 seconds, energized for 1.1 seconds (
(reversal) and 0.7 seconds of power cutoff are repeated. In this case, the forward and reverse rotational speeds of the agitator 9 are set to the "standard" speed of 72 rpm. in this way,
We try to prevent fabric damage as much as possible by quickly adapting the water flow to a strength that matches the condition of the fabric.

[0035] When the above-mentioned stirring by weak, standard, or strong water flow has been performed for a predetermined period of time, the microcomputer 13 stops the motor 3 again, and the motor 3 is brought into a stopped state (time T7). In this case as well, the light emitting diode 19 blinks and emits light. In this stopped state, the microcomputer 13 reads the transmittance signal S29 from the transmittance sensor 29 and stores it in the RAM as detection data A3. Thereby, as shown in FIG. 11, the microcomputer 13 determines that if the difference between the detection data A2 and A3 is small, the amount of detergent is small, and if the difference is large, the amount of detergent is large. , this is stored in the RAM as detergent amount data D5.

Thereafter, the microcomputer 13 rotates the motor 3, that is, the agitator 9, with the water flow set as described above (time T8), and when a predetermined period of time has elapsed, brings it to a stopped state again (time T9). In this case as well, the light emitting diode 1
9 flashes and emits light. And microcomputer 1
3 reads the transmittance signal S29 of the transmittance sensor 29 and stores it in the RAM as detection data A4 (time T10).

As shown in FIG. 11, the microcomputer 13 calculates A3-A2=ΔV1 from the detected data A2, A3, and A4 at times T5, T8, and T10.
, ΔV1' and A4-A3=ΔV2, ΔV2' are calculated, and since the difference between ΔV1 and ΔV2 is large, in this case, the quality of the dirt is determined to be mud-based, for example, and ΔV1' and ΔV2' are calculated.
In this case, the quality of the stain is determined to be oil-based, for example. That is, in the case of mud-based dirt, the dirt falls off quickly, so the rate of change between detected data A2 and A3 is large, and the ratio is large to the rate of change between subsequent detected data A3 and A4. On the other hand, in the case of oil-based stains, the degree of removal of the stain is gradual, so there is not much difference between the rate of change between detected data A2 and A3 and the rate of change between detected data A3 and A4. There is no difference. In this way, the microcomputer 13 calculates the dirt quality data D4 and stores it in the RAM.

As described above, the RAM of the microcomputer 13 stores cloth amount data D1 and cloth quality data D2 as cloth conditions, stain amount data D3 and stain quality data D4 as stain conditions, and also stores detergent Detergent amount data D5 and detergent quality data D6 are stored as conditions, and in addition, water temperature data D7 is stored. And these data D1, D2, D3, D4, D5, D6
and D7, as shown in FIG.
Units I1, I2, I3, I4, I5, of input layer I of
Neurocontrol is performed by inputting to I6 and I7, respectively.

Now, an outline of the theory of the neural network used for neural control will be described with reference to FIGS. 15 to 18.

[0040] A neural network simulates a biological neural network, and is a circuit consisting of units and links, as shown in FIG. Unit j has input/output characteristics Fj (Uj) as shown in FIG. Here, Fj is a sigmoid function, for example,

[0041]

[Math 1]

It is expressed by the formula: The output Vj of unit j is expressed by the following equation.

[0043]

[Math 2]

Here, Vi is the output of another unit i,
Wji is a weighting coefficient indicating the degree of influence of the output of unit i on unit j, and θj is a threshold value.

[0045] Neural nets are classified into fully connected types, hierarchical types, and intermediate types depending on the way links are connected.Here, an example of the hierarchical type will be shown.

FIG. 17 shows a three-layer neural network, in which units are arranged to form three layers called an input layer, a middle layer, and an output layer.
, j and k. In this neural network, signals are transmitted in one direction from the input layer through the intermediate layer to the output layer. In this case, a weighting coefficient Wji is set for the link from the input layer to the hidden layer,
A weighting coefficient Wkj is set for the link from the intermediate layer to the output layer. In this way, a neural network is composed of a large number of units that are simple arithmetic elements, and each unit outputs a large output when the sum of inputs from other units exceeds a threshold value.

[0047] Characteristics of neural networks include learning ability, high speed, and noise resistance.

Learning of the neural network is performed by adjusting the weighting coefficients of the links so that a desired output pattern (teacher pattern) can be obtained for the input pattern (example). In this case, the weighting factors are initially set randomly. After learning, the neural network has a plurality of learned pairs of input patterns and output patterns associated with each other, and can obtain a desired output pattern by analogy with input patterns other than the learned ones.

In this case, a backpropagation method is an example of a neural network learning method. In this backpropagation method, weighting coefficients are adjusted using an error function between a desired output pattern (teacher pattern) and an actual output pattern. In the neural network shown in FIG. 17, the error function E is defined as follows.

[0050]

[Math 3]

Here, Tk and Vk are teacher data (desired output data) and actual output data of unit k in the output layer, respectively. The output data Vk is expressed by the following equation.

[0052]

[Math 4]

[0053] In the backpropagation method,
The amount of modification of the weighting coefficient is calculated, and the modification is repeated until the amount becomes equal to or less than a certain value. In other words, the weighting coefficient correction amount ΔWk
j and ΔWji are

[0054]

[Math 5]

It can be found as follows. however,

[0056]

[Math 6]

[0057] Here, η is a constant determined from the balance between weighting coefficient correction speed and calculation stability. From these correction amounts ΔWkj and ΔWji,

00
58]

[Math 7]

The new weighting coefficients Wkj′ and Wji
′ and transfers learning to the next teacher pattern.

FIG. 18 shows the learning procedure. That is,
When learning about N pairs of input patterns and teacher patterns, first, the correction amount ΔWk is determined for "input pattern 1".
j, ΔWji are calculated and the weighting coefficients are corrected. next,"
Input pattern 2'', the correction amounts ΔWkj and ΔWji are calculated, and the weighting coefficients are corrected. Thereafter, the weighting coefficients are modified in the same manner up to "input pattern N." Then, if the correction amounts ΔWkj and ΔWji are not below a certain value, correction is repeated again from "input pattern 1",
If the value is below a certain value, learning ends.

[0061] The neural network described above can be constructed with hardware using a neurochip called a neuroprocessor, or it can be constructed in a simulated manner with software using a microcomputer. Yes, and in any case, it performs the function of a neural network.

The operation of the neural network 45 of this embodiment based on the above-mentioned theory will now be explained.

As shown in FIG. 1, cloth amount data D1, cloth quality data D2, dirt amount data D3, soil quality data D4, detergent amount data D5, detergent quality data D6, and water temperature data D7.
are input to each unit I1, I2, I3, I4, I5, I of the input layer I of the neural network 45 as a 4-bit signal, respectively.
6 and I7, and takes on any value of 16 categories from "1" to "16". Further, the units K1 to K3 of the output layer K of the neural network 45 output water flow data O1, washing time data O2, and rotation speed data O3 of the agitator 9 as output data, and these are also respectively output. It is output as a 4-bit signal.

In this case, the water flow data O1 is as shown in FIG.
), as shown in "1" (0.75 second energization and 1.35 seconds
It takes one of 16 values from ``16'' (1.1 second energization and 0.6 second energization). In addition, the washing time data O2 is divided into "washing process" and "rinsing process" as shown in FIG. "16" (
15 minutes), and for the ``rinsing process'', take a value from one of the 16 categories from ``1'' (5 minutes) to ``16'' (12.5 minutes). It has become. As shown in FIG. 12(c), the rotation speed data O3 ranges from "1" (30 rpm) to "16" (120 rpm).
It takes one of 16 values up to (rpm).

The calculations of the neural network 45 will now be explained.

Now, the input data D1, D which is the input pattern
When the values of 2, . . . , D7 are respectively UI1, UI2, . . . , UI7, the units I1, I2, . That is, VI1=UI1, VI2=UI2
,..., VI7=UI7.

Units J1, J2,..., J5 of middle layer J
For example, taking unit J1 as an example, its input UJ1 is

[0068]

[Math. 8]

[0069] In this case, WJ1I1 is the weighting coefficient from unit I1 to unit J1, WJ1I2 is the weighting coefficient from unit I2 to unit J1, ..., WJ1I
7 is a weighting coefficient from unit I7 to unit J1, and θJ1 is a threshold value. And unit J1 is
Using the above UJ1 as input, calculate the sigmoid function F,
The result is set as output VJ1. That is,

[0070]

[Math. 9]

[0071] The above also applies to units J2 to J5.

Regarding the units K1 to K3 of the output layer K, taking unit K1 as an example, its input UK1
teeth,

[0073]

[Math. 10]

[0074] In this case, WK1J1 is the weighting coefficient from unit J1 to unit K1, WK1J2 is the weighting coefficient from unit J2 to unit K1, ..., WK1J
5 is a weighting coefficient from unit J5 to unit K1, and θK1 is a threshold value. And unit K1 is
Using the above UK1 as input, calculate the Sigmond function F,
The result is set as output VK1. That is,

[0075]

[Math. 11]

[0076] The above also applies to units K2 and K3.

The weighting coefficient W and the threshold value θ are each represented by, for example, 4 bits, and can be positive, 0, or negative. For example, "0" is "0000", "1" is "
0001" and "-1" are represented by "1111". That is, the most significant bit is a negative sign bit. Further, the multiplication result (WV) of the weighting coefficient W and the output V takes the upper 5 bits, and the most significant bit is a negative sign bit. Further, the input U is represented by 8 bits, and the most significant bit is a negative sign bit. The output V is represented by 4 bits and takes a positive (or 0) value.

Regarding the threshold value θ, as shown in FIG. 14, a unit whose output is always "1" is set in the input layer I and the intermediate layer J, and the weighting coefficient of the link from that unit is set as θK. , θJ, the actual weighting coefficient WKJ,
It can be handled in exactly the same way as WJI. here,
The output "1" represents the maximum value that the output of the unit can take, that is, the maximum value of the output of the sigmoid function F, and is "16" in the above example. Like the weighting coefficient W, the threshold value θ can be positive, 0, or negative, but the number of bits can be different.

In the above case, one of each of the input layer I and the intermediate layer J of the neural network 45 may be set as the unit that takes the output "1", but it is also possible to set a new unit. Good too.

Learning of the neural network 45 is mainly performed at the product development stage. In this case, it is not necessary to learn all possible input patterns, and it is sufficient to learn about 20 patterns, for example. That is, FIG.
In this case, N is "20". If the weighting coefficient W and the threshold value θ are determined as a result of learning, the weighting coefficient W and the threshold value θ may be the same for the same model for mass production.

In this way, the neural network 45 outputs water flow data O1 and time data O2.
and rotation speed data 03 are calculated, and based on this result, the microcomputer 13 calculates "1" of the water flow data O1.
” to “16”, and select “1” to “16” for “washing process” of time data O2.
Select one of the values and set the rotation speed data O.
3 is selected from "1" to "16" to execute the stirring or "washing" process after time T10 in FIG.

FIG. 13 shows the rotation state of the agitator 9 during this "washing" process, where R is the rotation speed selected from FIG. 12(c), and S1 is the rotation speed selected from FIG. 12(b). S2 is the power-off time between forward and reverse rotations selected from S2.

Thereafter, when the selected washing time period ends (time T11), the microcomputer 1
3, the power to the motor 3 is cut off, thus ending the "washing" process.

Now, when the above-mentioned "washing" process is completed (time T11), the microcomputer 13
The process moves to a "rinsing" process as shown in FIG. This "rinsing" process includes "draining" which opens the drain valve 5 by energizing the electromagnet 35 via the drive circuit 34, and rotating the inner tank 7 by energizing the motor 3 with the drain valve 5 open. "Dehydration", "Water supply" which opens the water supply valve 11, "First rinse" which performs "rinsing" similar to the above-mentioned stirring, and "Drainage", "Dehydration", "Water supply" and "Rinsing" similar to the above. This consists of a "second rinse" in which the process is performed. In this case, "water supply" is set to the water level set during the "washing" process, and "rinsing" is set to the water flow and rotation speed set during the "washing" process.

[0085] In the "first rinse", the "rinsing" time H1 is set to a certain time (for example, 3 minutes), and a preliminary rinse is performed, and in the "second rinse", the "rinsing" The time H2 is set to be the time H1 selected from the "rinsing process" in FIG. 12(b) minus the time H1, and the water injection rinsing is performed. That is, the time H for the "rinsing process" shown in FIG. 12(b) is the total time (H= H1+H2).

Next, upon completion of the "rinsing" process, the microcomputer 13 shifts to the "dehydration" process as shown in FIG. In this "dehydration" process, "drainage" and "dehydration" similar to those in the "rinsing" process are performed.

FIG. 21 shows changes in the permeability of the dehydrating liquid during the "dehydration" process, and this change in permeability is detected by the permeability sensor 20. That is, "rinsing"
When the process is completed (time T12) and "draining" is started, the permeability remains as long as the rinsing liquid remains in the outer tank 2.
This shows the permeability during the "rinsing" process, but when the "draining" process is completed (time T13), the permeability sensor 20 detects the permeability of air because the rinsing liquid has been discharged and is no longer present. . After that, when "dehydration" starts, the rinsing liquid contained in the laundry is shaken off and accumulates at the bottom of the outer tub 2, so the permeability sensor 20 detects the permeation of the rinsing liquid during the ``rinsing'' process. It will detect the degree.

Thereafter, as the "dehydration" progresses, the rinsing liquid shaken off from the laundry is discharged and disappears, so the permeability detected by the permeability sensor 20 increases toward the permeability of air. Then, when the permeability becomes a difference ΔV3 from the permeability of air (time T14), the microcomputer 13 ends the "dehydration" process after a certain period of time TC (time T15).

In this case, as shown in FIG. 22, when the amount of rinsing liquid contained in the laundry is standard, the difference ΔV3 is reached at time T14 as shown by the solid line, but when it is small, the difference ΔV3 is reached at time T14 as shown by the dashed line. The shorter time T14a
When the difference is large, the difference ΔV3 is reached at time T14b, which is longer than time T14, as shown by the broken line. That is, the time of the "spin" process is adjusted and set depending on the amount of rinsing liquid contained in the laundry.

According to this embodiment, the following effects can be obtained.

That is, as the state of laundry etc., cloth condition is cloth amount data D1 and cloth quality data D2, stain condition is stain amount data D3 and stain quality data D4, detergent condition is detergent amount data D5 and detergent quality data D6, Then, the water temperature data D7 is transferred to the neural network 4 of the microcomputer 13.
5, the strength of the washing water flow (i.e.,
energization time and de-energization time of motor 3 and rotation speed of stirring body 9)
In addition to determining the "washing" time and "washing time"
Since the rinsing time is determined, it is possible to obtain the washing effect and the rinsing effect corresponding to various conditions, and it is possible to prevent fabric stains due to over-washing or over-rinsing, and also to prevent fabric stains due to insufficient washing or rinsing. The trouble of re-washing or re-rinsing due to shortage can be avoided.

Furthermore, during the period T4-T5 during which the stirring body 9 is stopped in order to obtain the detection data A2, A3, and A4,
Between 7 and T8 and between T9 and T10, the light emitting diode 19 with the display ``neurosensor'' blinks and emits light, so that the washing operation stops midway and the user misunderstands that there is a malfunction. It has the advantage of not causing any damage.

Furthermore, in the ``rinsing'' step, the ``rinsing'' time of the ``second rinsing'' water injection rinsing is set by neurocontrol, so that an optimal rinsing effect can be obtained.

In the "dehydration" process, the amount of rinsing liquid contained in the laundry is determined based on the output of the permeability sensor 20, and the "dehydration" time is varied, so that the optimal A dehydrating effect can be obtained.

In the above embodiment, the sigmoid function F is obtained by calculation, but the calculation result may be obtained by a matrix as in the second embodiment of the present invention shown in FIG.

That is, FIG. 23 is an example of a matrix representing the sigmoid function F, in which the vertical axis represents the output V and the horizontal axis represents the input U. In this case, for example, input U is 0 (U=
0), the output V is 9 (V=9).

In the above embodiment, the strength of the washing water flow is determined by both the energization time of the agitator 9 in the forward and reverse directions, the power-off time between the forward and reverse directions, and the rotational speed of the agitator 9. However, one of them may be fixed and the other may be changed.

In each of the above embodiments, the input data of the state of laundry etc. is cloth amount data D1, cloth quality data D2, dirt amount data D3, soil quality data D4, and detergent amount data D5.
, the detergent quality data D6 and the water temperature data D7 are input into the neural network 45, but it is of course possible to input conditions other than these as data, and at least the cloth amount data D1 which is the cloth condition. , fabric quality data D2, dirt amount data D4 as dirt conditions, and dirt quality data D4, sufficient results for neurocontrol can be obtained.

Further, in each of the above embodiments, neurocontrol is performed in the "standard" course of the washing courses, but neurocontrol may be performed in the same manner in other washing courses. .

[0100] In addition, the present invention is not limited to the embodiments described above and shown in the drawings, but can of course be implemented with appropriate modifications within the scope of the gist.

[0101]

[Effects of the Invention] As explained above, the fully automatic washing machine of the present invention performs neuro-control of the washing operation based on the input data from the sensor that detects the condition of the laundry, etc., so that the optimum washing effect can be achieved. You can expect it.

[Brief explanation of drawings]

FIG. 1 is a diagram for explaining the outline of neurocontrol showing a first embodiment of the present invention.

[Figure 2] Overall vertical cross-sectional view

[Figure 3] Block diagram showing electrical configuration

[Figure 4] Top view of the operation panel

[Figure 5] Longitudinal cross-sectional view of the transmittance sensor

[Figure 6] Longitudinal cross-sectional view of temperature sensor

[Figure 7] Electrical wiring diagram of capacitive sensor

[Figure 8] Waveform diagram for explaining the action of the capacitive sensor

[Figure 9] Time chart for explaining the action of the washing process

[Figure 10] Diagram showing input/output data and process flow

[Figure 11] Transmittance characteristic diagram

[Figure 12] Diagram showing output results of water flow, time, and rotation speed

[
Figure 13: Time chart showing the rotational state of the stirring body

[Figure 1
4] Diagram explaining the effect of determining the threshold value

[Figure 15] Basic diagram of a neural network to explain the principle of neurocontrol

[Figure 16] Diagram showing the sigmoid function

[Figure 17] Diagram showing the same three-layer neural network

[Figure 18
] Diagram showing the learning procedure

[Figure 19] Process explanatory diagram of the rinse process

[Figure 20] Process explanatory diagram of the dehydration process

[Figure 21] Transmittance characteristic diagram

[Figure 22] Dehydration degree detection action diagram

FIG. 23: Matrix showing the second embodiment of the present invention

[Explanation of symbols]

In the drawing, 2 is an inner tank, 3 is a motor, 5 is a drain valve, 7 is an outer tank, 9 is an agitator, 11 is a water supply valve, 13 is a microcomputer (control means), 14 is an operation panel, and 20 is a transparency 24 is a temperature sensor, 28 is a rotational speed sensor, 29 is a capacitance sensor, and 45 is a neural network.

Claims (1)

[Claims] Claim 1: The water flow intensity and washing time of a washing operation are determined by neurocontrol in which weighting coefficients and threshold values learned in advance are set based on input data from a sensor that detects the condition of laundry, etc. A fully automatic washing machine characterized by comprising a control means for causing the washing machine to execute the washing machine.