Neural network model training method, washing machine and control method

Abstract

The invention discloses a training method of a neural network model, a washing machine and a control method. The training method of the neural network model comprises the following steps: converting current in a data set into q-axis current and d-axis current; respectively carrying out FFT (Fast Fourier Transform) to obtain a current frequency spectrum; determining the rotating speed of the motor, calculating the mechanical rotating frequency of the motor, and extracting harmonic amplitudes of a double frequency and a double frequency in a current spectrum; taking the extracted harmonic amplitude and the motor rotating speed as the input of a neural network model, wherein the output of the neural network model comprises weight, eccentricity and diagonal eccentricity; and training the model according to the eccentricity, diagonal eccentricity and weight output by the neural network model and corresponding actual values. The model is trained through harmonic amplitudes of one time and two times of frequencies in a current spectrum, and the obtained model can predict eccentricity and diagonal eccentricity. Meanwhile, the change characteristic of the extracted harmonic amplitude is more obvious than that of the direct use of the current, the model is easier to train, and the accuracy of model output is improved.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence technology, and in particular relates to a training method for a neural network model, a washing machine and a control method. Background Art

[0002] Currently, the weight, eccentricity and diagonal eccentricity of a washing machine load are mainly detected by setting sensors.

[0003] With the development of artificial intelligence technology in recent years, the weight of the load has also been predicted through neural network models. The training of neural network models is mainly carried out through parameters such as motor speed and current.

[0004] Directly using the motor drive current as input for the neural network model requires the model to identify the current's changing characteristics. This results in high complexity, long training times, poor generalization, and large errors in the resulting results, making it difficult to achieve the desired results. Most importantly, existing neural network models cannot simultaneously predict both eccentricity and diagonal eccentricity.

[0005] The above information disclosed in this background technology is only used to increase the understanding of the background technology of this application. Therefore, it may contain information that does not constitute the prior art known to ordinary technicians in this field. Summary of the Invention

[0006] The present invention proposes a training method for a neural network model, a washing machine and a control method to solve the technical problems that the existing washing machine uses parameters such as motor speed and current as neural network model inputs, resulting in large prediction errors and the inability to simultaneously predict eccentricity and diagonal eccentricity.

[0007] In order to achieve the above invention / design purpose, the present invention adopts the following technical solutions: A neural network model training method for a washing machine, the method comprising: Acquire a data set; the data set includes the motor speed and the current of the motor UV phase line when several different loads, eccentric loads, and diagonal eccentric loads are placed in the washing machine drum 、 ; The current of the motor UV phase line 、 Convert the q-axis current and d-axis current into the q-axis current; divide the q-axis current and d-axis current according to the set sampling frequency and the set sliding window division time to obtain several frames of q-axis current signals and d-axis current signal , perform FFT transformation respectively to obtain the q-axis current spectrum and d-axis current spectrum ; Wherein, x represents the xth sliding window split time, and x is a natural number; Determine the motor speed corresponding to the xth sliding window split time , calculate the motor mechanical rotation frequency ; Extract the q-axis current spectrum and d-axis current spectrum middle The harmonic amplitude and Harmonic amplitude; extracted harmonic amplitude and motor speed As input to a neural network model, the output of the neural network model includes weight, eccentricity, and diagonal eccentricity; The neural network model is trained using the eccentricity, diagonal eccentricity and weight output by the neural network model and corresponding actual values to obtain a trained neural network model.

[0008] The neural network model training method for the washing machine described above divides the extracted harmonic amplitude by the fundamental amplitude to obtain the harmonic energy. , , l=1, 2; based on the harmonic energy and motor speed as the input of the neural network model.

[0009] As described above, the neural network model training method for a washing machine has the following formula for the jth neuron in the i-th hidden layer of the neural network model: ;in, is the output of the jth neuron node in the previous layer, for The weight of the jth neuron in the layer, for The bias of the jth neuron in the layer, i is the layer where the neuron is located, j is the number of neurons in each layer, for The total number of neurons in the layer; The output of the neuron is obtained through the activation function , activation function is ReLU; Mean absolute error loss function: ,in, is the actual value, is the output value of the neural network model, M is the number of samples, and m is the mth sample; substitute the actual value and the output value of the neural network model into the loss function to calculate the loss value of the current neural network model; Using the chain rule starting from the output layer, calculate the gradient of the loss function with respect to each weight and bias. The layers are calculated as follows: ; in, For the layer weight vector, For the layer bias vector; According to the calculated gradient, the weights and biases of the neural network model are updated: ; in, is the learning rate; Iterate the training until the loss function converges or the predetermined number of training rounds is reached to obtain a trained neural network model.

[0010] In the neural network model training method for a washing machine as described above, the motor speed in the data set includes at least two different motor speeds and an intermediate speed between the two different motor speeds.

[0011] A method for controlling a washing machine, comprising: Start to become dehydrated; Controlling the motor speed to reach a first speed; Collect the motor speed and the current of the motor UV phase line 、 , the current of the motor UV phase line 、 Convert to q-axis current and d-axis current, perform FFT transformation on the q-axis current and d-axis current respectively, and obtain the q-axis current spectrum and d-axis current spectrum ; Determine the corresponding motor speed , calculate the motor mechanical rotation frequency ; ; Extract the q-axis current spectrum and d-axis current spectrum middle The harmonic amplitude and Harmonic amplitude; extracted harmonic amplitude and motor speed As input to a trained neural network model, the trained neural network model outputs weight, eccentricity, and diagonal eccentricity; and controlling the washing machine based on the weight, eccentricity, and diagonal eccentricity; The trained neural network model is a neural network model trained using the above-mentioned training method.

[0012] The control method of the washing machine as described above determines whether it is qualified based on the eccentricity and diagonal eccentricity output by the trained neural network model. If so, the motor is controlled to speed up to the second speed for dehydration; if not, the motor is controlled to run at the first speed.

[0013] According to the control method of the washing machine described above, when the motor speed is at the second speed for dehydration, the eccentricity is monitored, and when the eccentricity is qualified, dehydration is continued until the dehydration end condition is reached; when the eccentricity is unqualified, the motor is controlled to run at the first speed.

[0014] In the control method of the washing machine as described above, the method for monitoring the eccentricity is: The eccentricity output by the trained neural network model is used for monitoring.

[0015] In the control method of the washing machine as described above, the method for monitoring the eccentricity is: The eccentricity output by the second neural network model is used for monitoring; The second neural network model is trained by the following method: Acquire a data set, wherein the data set includes motor speed, total vector current Is, and total vector voltage Vs of the motor detected when several different eccentric loads are placed in the drum of the drum washing machine; The motor speed, the total vector current Is and the total vector voltage Vs of the motor are used as inputs of a second neural network model, wherein the output of the second neural network model includes eccentricity; The motor speed in the second neural network model data set is greater than the motor speed in the neural network model data set.

[0016] In the control method for the washing machine described above, the formula for the jth neuron in the i-th hidden layer of the second neural network model is: ; in, is the output of the jth neuron node in the previous layer, for The weight of the jth neuron in the layer, for The bias of the jth neuron in the layer, i is the layer where the neuron is located, j is the number of neurons in each layer, for The total number of neurons in the layer; The output of the neuron is obtained through the activation function , activation function is ReLU; mean absolute error loss function: ,in, is the actual value, is the output value of the neural network model, M is the number of samples, and m is the mth sample; substitute the actual value and the output value of the neural network model into the loss function to calculate the loss value of the current neural network model; Using the chain rule starting from the output layer, calculate the gradient of the loss function with respect to each weight and bias. The layers are calculated as follows: ; in, For the layer weight vector, For the layer bias vector; According to the calculated gradient, the weights and biases of the second neural network model are updated: ; in, is the learning rate; The training is iterated until the loss function converges or the predetermined number of training rounds is reached, and a trained second neural network model is obtained.

[0017] Compared with the prior art, the advantages and positive effects of the present invention are as follows: The neural network model training method for a washing machine: a data set is obtained, which includes the motor speed and the current of the motor UV phase line when several different loads, eccentric loads, and diagonal eccentric loads are placed in the washing machine drum. 、 ; The current of the motor UV phase line 、 The method converts the current into q-axis and d-axis currents; performs FFT transformations on the q-axis and d-axis currents to obtain current spectra; determines the motor speed, calculates the motor mechanical rotational frequency, and extracts the harmonic amplitudes at 1x and 2x frequencies from the current spectrum; uses the extracted harmonic amplitudes and motor speed as inputs to a neural network model, whose outputs include weight, eccentricity, and diagonal eccentricity; and trains the neural network model using the eccentricity, diagonal eccentricity, and weight outputs from the neural network model and the corresponding actual values, resulting in a trained neural network model. The eccentricity spectrum shows a significant increase in the 1x frequency component, a slight increase in the 2x frequency component, and a significant increase in the 2x frequency component in the diagonal spectrum. Therefore, by performing FFT transformations on the q-axis and d-axis currents to obtain current spectra, extracting the harmonic amplitudes at 1x and 2x frequencies from the current spectrum, and training the neural network model, the trained neural network model can predict weight, eccentricity, and diagonal eccentricity. Furthermore, using the extracted harmonic amplitudes provides more distinct variations than directly using the current, making it easier to train the neural network model and improving the accuracy of the model output.

[0018] The control method of the washing machine can predict the weight, eccentricity and diagonal eccentricity of the load with a trained neural network model. At the same time, the use of the extracted harmonic amplitude has obvious variation characteristics compared to directly using the current, which makes it easier to train the neural network model and improve the accuracy of the model output, thereby making the control of the washing machine more precise.

[0019] Other features and advantages of the present invention will become more apparent after reading the detailed description of the present invention in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0021] Figure 1 It is a schematic diagram showing that the load is evenly arranged in the cylinder according to a specific embodiment of the present invention.

[0022] Figure 2 yes Figure 1 Top view of area 1.

[0023] Figure 3 It is a schematic diagram of a load arranged in an eccentric state according to a specific embodiment of the present invention.

[0024] Figure 4 yes Figure 3 Top view of area 1.

[0025] Figure 5 It is a schematic diagram of a load arrangement in a diagonally eccentric state according to a specific embodiment of the present invention.

[0026] Figure 6 yes Figure 5 Top view of area 1.

[0027] Figure 7 yes Figure 5 Top view of area 2.

[0028] Figure 8 This is the motor speed change diagram.

[0029] Figure 9 It is a model diagram of the first neural network model in a specific embodiment of the present invention.

[0030] Figure 10 4 is a flow chart of a washing machine control method according to a specific embodiment of the present invention.

[0031] Figure 11 It is a model diagram of the second neural network model in a specific embodiment of the present invention.

[0032] Figure 12 It is a flow chart of a control method for a washing machine according to another specific embodiment of the present invention. DETAILED DESCRIPTION

[0033] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0034] In the description of the present invention, it should be understood that the terms "center", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be understood as a limitation on the present invention.

[0035] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "installed," "connected," and "connected" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integral connections. A person of ordinary skill in the art will understand the specific meanings of the above terms in the present invention in specific circumstances. In the description of the embodiments, specific features, structures, materials, or characteristics may be combined in any appropriate manner in any one or more embodiments or examples.

[0036] The terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the quantity of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the features.

[0037] In the description of the present invention, unless otherwise specified, "plurality" means two or more.

[0038] Washing machines are weighed to determine the weight of the laundry. This allows the appropriate water level, wash time, and detergent dosage to be calculated based on the weight and type of laundry, ensuring thorough cleaning while minimizing water and detergent consumption. Therefore, washing machine weighing detection can provide more efficient and reliable cleaning. Washing machine eccentricity and diagonal eccentricity detection can identify the distribution of laundry within the washing machine, avoiding vibration and noise caused by uneven distribution of laundry, reducing damage to internal components, and extending the machine's service life. Eccentricity detection during high-speed spin cycles can help address the problem of poorly permeable laundry, which can cause the drum to collide due to sudden changes in balance caused by the release of water during the spin cycle.

[0039] A neural network model training method for a washing machine, comprising: Get the dataset.

[0040] The data set includes the motor speed and the current of the motor UV phase line when several different loads, eccentric loads, and diagonal eccentric loads are placed in the washing machine drum. 、 .

[0041] exist Figure 1-Figure 7 In the example, the cylinder is divided into three areas according to 120 degrees, and the counterweights are placed according to the areas.

[0042] In some embodiments, the load is increased by 2 x 500g per area at a time, and the counterweights are placed from left to right according to the working conditions. For example, for a load of 1000g, 500g counterweights are placed in positions 1 and 2 in all areas; for a load of 2000g, 500g counterweights are placed in positions 1, 2, 3, and 4 in all areas; when the load exceeds 4000g, the new counterweights are stacked on top of the counterweights in positions 1 and 2, and so on.

[0043] The eccentric block (red block in the picture) is fixedly placed at position 4 in each area, increasing in units of 200g.

[0044] When measuring load eccentricity, first configure the counterweights and measure the corresponding eccentric weights in each area. Place the eccentric weights on the load block in position 4, in increments of 200g.

[0045] When testing diagonal eccentricity, in test area 1, place the diagonal eccentric at position 2 and align it with the outer cylinder wall. Place an eccentric block of equal weight at position 7 in area 2 and align it with the inner cylinder wall. Repeat the same process for other areas.

[0046] The current of the motor UV phase line 、 Converted into q-axis current and d-axis current.

[0047] The current of the motor UV phase line is collected through the sampling resistor 、 , converted to Axis coordinate system 、 .

[0048] .

[0049] Will Axis coordinate system 、 Convert to dq axis coordinate system 、 .

[0050] ; in, is the electrical angle of the motor rotor.

[0051] The motor speeds in the data set include at least two different motor speeds and an intermediate speed between the two different motor speeds.

[0052] exist Figure 8 In the example, the motor speed is controlled to be 90 rpm and then data is collected after it stabilizes. Data is collected during the process of accelerating to 120 rpm and then after it stabilizes at 120 rpm. Data is then collected during the process of decelerating to 90 rpm and then after it stabilizes at 90 rpm.

[0053] Divide the sampling data set: Divide the collected data into a training set (80%), a validation set (10%), and a test set (10%). The training set is used for model training, the validation set is used to adjust the model hyperparameters, and the test set is used to evaluate the final performance of the model.

[0054] Preprocess the data set and perform operations such as normalization and standardization on the data to make the data have similar scale and distribution to accelerate model convergence.

[0055] Data feature extraction: filtering, sliding window segmentation and FFT are used to extract the harmonic spectrum characteristics of the q-axis and d-axis currents.

[0056] (1) Perform a first-order low-pass filter on the collected data to remove the noise data.

[0057] (2) The sampling frequency is the set sampling frequency, and the sliding window split time is the set sliding window split time.

[0058] (3) The q-axis current and d-axis current are divided according to the set sampling frequency and the set sliding window division time to obtain several frames of q-axis current signals. and d-axis current signal , perform FFT transformation respectively to obtain the q-axis current spectrum and d-axis current spectrum ; Where x represents the x-th sliding window split time, and x is a natural number.

[0059] ;

[0060] .

[0061] Fundamental frequency identification: Determine the motor speed corresponding to the x-th sliding window split time , calculate the motor mechanical rotation frequency .

[0062] Harmonic location: Extracting the q-axis current spectrum and d-axis current spectrum middle The harmonic amplitude and harmonic amplitude.

[0063] : The main characteristic frequency of out-of-center (OOB).

[0064] : eigenfrequency of diagonal off-centering (DOOB).

[0065] OoB spectrum performance: Significantly increased in weight, Slightly increased in weight.

[0066] Diagonal Offset (DOOB) spectrum performance: Significantly increased portion size.

[0067] The extracted harmonic amplitude and motor speed As the input of the neural network model, the output of the neural network model includes weight, eccentricity and diagonal eccentricity.

[0068] The neural network model is trained using the eccentricity, diagonal eccentricity and weight output by the neural network model and the corresponding actual values to obtain a trained neural network model.

[0069] In some embodiments, the extracted harmonic amplitude is divided by the fundamental amplitude to obtain the harmonic energy , , l=1, 2; based on harmonic energy and motor speed As input to the neural network model.

[0070] Dividing the amplitude of each harmonic by the amplitude of the fundamental wave gives the relative harmonic energy (in percentage form), which can eliminate the influence of motor power differences.

[0071] The harmonic amplitude of the extracted data is used as the input of the neural network, which is easier to fit the neural network model than directly using the change characteristics of the current.

[0072] exist Figure 9 In the example, the input layer of the first neural network model consists of 3 neurons, which correspond to the rotational speed, q-axis current harmonic amplitude, and d-axis current harmonic amplitude of the collected sample data; the output layer consists of 3 neurons, which correspond to the eccentricity (OOB), diagonal eccentricity (DOOB), and weight of the sample calibration data.

[0073] The formula for the jth neuron in the i-th layer of the hidden layer of the neural network model is: .

[0074] in, is the output of the jth neuron node in the previous layer, for The weight of the jth neuron in the layer, for The bias of the jth neuron in the layer, i is the layer where the neuron is located, j is the number of neurons in each layer, for The total number of neurons in the layer.

[0075] The output of the neuron is obtained through the activation function , activation function is ReLU.

[0076] Mean absolute error loss function: ,in, is the actual value, is the output value of the neural network model, M is the number of samples, and m is the mth sample.

[0077] Substitute the actual value and the output value of the neural network model into the loss function to calculate the loss value of the current neural network model, which reflects the degree of difference between the model prediction and the actual situation.

[0078] Using the chain rule starting from the output layer, calculate the gradient of the loss function with respect to each weight and bias. The layers are calculated as follows: ; in, For the layer weight vector, For the layer bias vector; According to the calculated gradient, the weights and biases of the neural network model are updated using the gradient descent method as follows: ; in, is the learning rate.

[0079] Iterate training until the loss function converges or the predetermined number of training rounds is reached. Evaluate the model's performance on the validation set and adjust hyperparameters to prevent overfitting. Finally, evaluate the model's generalization ability on the test set.

[0080] Get the trained neural network model.

[0081] A control method for a washing machine: Start dehydrating.

[0082] The motor speed is controlled to reach a first speed.

[0083] In some embodiments, the first rotational speed is a distributed rotational speed.

[0084] Collect motor speed and current of motor UV phase line 、 , the current of the motor UV phase line 、 Convert to q-axis current and d-axis current, perform FFT transformation on q-axis current and d-axis current respectively, and obtain q-axis current spectrum and d-axis current spectrum .

[0085] Determine the corresponding motor speed , calculate the motor mechanical rotation frequency .

[0086] ; Extract q-axis current spectrum and d-axis current spectrum middle The harmonic amplitude and harmonic amplitude.

[0087] The extracted harmonic amplitude and motor speed As input to the trained neural network model, the trained neural network model outputs weight, eccentricity, and diagonal eccentricity.

[0088] The washing machine is controlled according to weight, eccentricity and diagonal eccentricity.

[0089] The trained neural network model is the first neural network model trained using the above training method.

[0090] Whether the eccentricity and diagonal eccentricity output by the trained first neural network model are qualified is judged. If so, the motor is controlled to speed up to the second speed for dehydration. If not, the motor is controlled to run at the first speed.

[0091] In some embodiments, the second speed is a high-speed dehydration speed.

[0092] When the motor speed is at the second speed for dehydration, the eccentricity is monitored, and when the eccentricity is qualified, dehydration is continued until the dehydration end condition is reached. When the eccentricity is unqualified, the motor is controlled to run at the first speed.

[0093] The methods for monitoring eccentricity are: The eccentricity of the output of the first trained neural network model is used for monitoring.

[0094] exist Figure 10 In an example, the control method of the washing machine includes: S1. Start dehydration.

[0095] S2. Control the motor speed to reach a first speed.

[0096] S3. Collect data, extract current harmonic amplitudes, and use the extracted current harmonic amplitudes and motor speed as inputs of the first neural network model.

[0097] S4. The first neural network model outputs weight, eccentricity and diagonal eccentricity.

[0098] S5. Determine whether the eccentricity and diagonal eccentricity are qualified. If qualified, proceed to step S6; otherwise, proceed to step S2.

[0099] S6. Control the motor to the second speed for dehydration.

[0100] S7. Monitoring eccentricity through the first neural network model.

[0101] S8. Determine whether the eccentricity is qualified. If so, proceed to step S9; otherwise, proceed to step S2.

[0102] S9: The dehydration end condition is met. If so, go to step S10; otherwise, go to step S7.

[0103] S10, dehydration ends.

[0104] In some embodiments, the method of monitoring eccentricity is: The eccentricity output by the second neural network model is used for monitoring.

[0105] The second neural network model is trained by the following method: Obtain a data set, which includes the motor speed, total vector current Is, and total vector voltage Vs detected when several different eccentric loads are placed in the drum of a drum washing machine; The motor speed, the total vector current Is and the total vector voltage Vs of the motor are used as inputs of the second neural network model, and the output of the second neural network model includes eccentricity.

[0106] The motor speed in the second neural network data set is greater than the motor speed in the neural network data set.

[0107] In some embodiments, the motor speed in the second neural network data set is 1200 rpm.

[0108] Divide the sampling data set: Divide the collected data into a training set (80%), a validation set (10%), and a test set (10%). The training set is used for model training, the validation set is used to adjust the model hyperparameters, and the test set is used to evaluate the final performance of the model.

[0109] Data preprocessing: Perform normalization and standardization on the data to make the data have similar scale and distribution to accelerate model convergence. exist Figure 11 In the example, the input layer of the second neural network model consists of 3 neurons, corresponding to the speed of the sample data, The total vector current Is and the total vector voltage Vs in the coordinate system; the output layer consists of 1 neuron, corresponding to the sample calibration data eccentricity (OOB).

[0110] The formula for the jth neuron in the i-th layer of the hidden layer of the second neural network model is: .

[0111] in, is the output of the jth neuron node in the previous layer, for The weight of the jth neuron in the layer, for The bias of the jth neuron in the layer, i is the layer where the neuron is located, j is the number of neurons in each layer, is the total number of neurons in the layer.

[0112] The output of the neuron is obtained through the activation function , activation function is ReLU.

[0113] Mean absolute error loss function: ,in, is the actual value, is the output value of the neural network model, M is the number of samples, and m is the mth sample.

[0114] Substitute the actual value and the output value of the neural network model into the loss function to calculate the loss value of the current neural network model, which reflects the degree of difference between the model prediction and the actual situation.

[0115] Using the chain rule starting from the output layer, calculate the gradient of the loss function with respect to each weight and bias. The layers are calculated as follows: ; in, For the layer weight vector, For the Layer bias vector.

[0116] Based on the calculated gradient, the weights and biases of the second neural network model are updated.

[0117] ;

[0118] in, is the learning rate.

[0119] Iterate training until the loss function converges or the predetermined number of training rounds is reached. Evaluate the model's performance on the validation set, adjust hyperparameters to prevent overfitting, and finally evaluate the model's generalization ability on the test set.

[0120] The trained second neural network model is obtained.

[0121] Before dehydration, the first neural network model is used to predict weight, eccentricity and diagonal eccentricity during the low-speed rotation of the drum. At this time, the amplitude-frequency characteristics of the d-axis and q-axis currents are more obvious, and the real-time requirements for detection are not high. Using the amplitude-frequency characteristics of the current as the input layer variable of the first neural network model is more suitable to improve the prediction accuracy. In the high-speed dehydration stage, due to the high real-time requirements for calculating OOB, The total vector current Is and the total vector voltage Vs in the coordinate system are used as the input layer of the feedforward neural network model to meet the real-time requirements.

[0122] According to the different states of the drum at high and low speeds, the use of two neural network models for detection can output higher detection accuracy while meeting the timeliness requirements.

[0123] exist Figure 12 In the example of , the control method of the washing machine is: S1. Start dehydration.

[0124] S2. Control the motor speed to reach a first speed.

[0125] S3. Collect data, extract current harmonic amplitudes, and use the extracted current harmonic amplitudes and motor speed as inputs of the first neural network model.

[0126] S4. The first neural network model outputs weight, eccentricity and diagonal eccentricity.

[0127] S5. Determine whether the eccentricity and diagonal eccentricity are qualified. If qualified, proceed to step S6; otherwise, proceed to step S2.

[0128] S6, controlling the motor to increase the speed to the second speed for dehydration.

[0129] S7. Monitoring eccentricity through a second neural network model.

[0130] S8. Determine whether the eccentricity is qualified. If so, proceed to step S9; otherwise, proceed to step S2.

[0131] S9: If the dehydration end condition is met, go to step S10; otherwise, go to step S7.

[0132] S10, dehydration ends.

[0133] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for a person skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to replace some of the technical features therein with equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions claimed to be protected by the present invention.

Claims

1. A neural network model training method for a washing machine, characterized in that: The method comprises: Acquire a data set; the data set includes the motor speed and the current of the motor UV phase line when several different loads, eccentric loads, and diagonal eccentric loads are placed in the washing machine drum 、 ; The current of the motor UV phase line 、 Convert the q-axis current and d-axis current into the q-axis current; divide the q-axis current and d-axis current according to the set sampling frequency and the set sliding window division time to obtain several frames of q-axis current signals and d-axis current signal , perform FFT transformation respectively to obtain the q-axis current spectrum and d-axis current spectrum ; Wherein, x represents the xth sliding window split time, and x is a natural number; Determine the motor speed corresponding to the xth sliding window split time , calculate the motor mechanical rotation frequency ; Extract the q-axis current spectrum and d-axis current spectrum middle The harmonic amplitude and The harmonic amplitude of The extracted harmonic amplitude and motor speed As input to a neural network model, the output of the neural network model includes weight, eccentricity, and diagonal eccentricity; The neural network model is trained using the eccentricity, diagonal eccentricity and weight output by the neural network model and corresponding actual values to obtain a trained neural network model.

2. The neural network model training method for a washing machine according to claim 1, characterized in that: Divide the extracted harmonic amplitude by the fundamental amplitude to obtain the harmonic energy , , l=1, 2; based on the harmonic energy and motor speed as the input of the neural network model.

3. The neural network model training method for a washing machine according to claim 1 or 2, characterized in that: The formula for the jth neuron in the i-th layer of the hidden layer of the neural network model is: ； in, is the output of the jth neuron node in the previous layer, for The weight of the jth neuron in the layer, for The bias of the jth neuron in the layer, i is the layer where the neuron is located, j is the number of neurons in each layer, for The total number of neurons in the layer; The output of the neuron is obtained through the activation function , activation function is ReLU; Mean absolute error loss function: ,in, is the actual value, is the output value of the neural network model, M is the number of samples, and m is the mth sample; Substitute the actual value and the output value of the neural network model into the loss function to calculate the loss value of the current neural network model; Using the chain rule starting from the output layer, calculate the gradient of the loss function with respect to each weight and bias. The layers are calculated as follows: ； in, For the layer weight vector, For the layer bias vector; According to the calculated gradient, the weights and biases of the neural network model are updated: ； in, is the learning rate; Iterate the training until the loss function converges or the predetermined number of training rounds is reached to obtain a trained neural network model.

4. The neural network model training method for a washing machine according to claim 1, characterized in that: The motor speed in the data set includes at least two different motor speeds and an intermediate speed between the two different motor speeds.

5. A method for controlling a washing machine, characterized in that: The method is: Start to become dehydrated; Controlling the motor speed to reach a first speed; Collect the motor speed and the current of the motor UV phase line 、 , the current of the motor UV phase line 、 Convert to q-axis current and d-axis current, perform FFT transformation on the q-axis current and d-axis current respectively, and obtain the q-axis current spectrum and d-axis current spectrum ; Determine the corresponding motor speed , calculate the motor mechanical rotation frequency ; ； Extract the q-axis current spectrum and d-axis current spectrum middle The harmonic amplitude and The harmonic amplitude of The extracted harmonic amplitude and motor speed As input to a trained neural network model, the trained neural network model outputs weight, eccentricity, and diagonal eccentricity; controlling the washing machine according to the weight, eccentricity and diagonal eccentricity; The trained neural network model is a neural network model trained by the training method of any one of claims 1-4.

6. The control method of the washing machine according to claim 5, characterized in that: Whether the eccentricity and diagonal eccentricity output by the trained neural network model are qualified is judged. If so, the motor is controlled to speed up to the second speed for dehydration. If not, the motor is controlled to run at the first speed.

7. The control method of the washing machine according to claim 6, characterized in that: When the motor speed is at the second speed for dehydration, the eccentricity is monitored, and when the eccentricity is qualified, dehydration is continued until a dehydration end condition is reached; when the eccentricity is unqualified, the motor is controlled to run at the first speed.

8. The control method of the washing machine according to claim 7, characterized in that: The method for monitoring the eccentricity is: The eccentricity output by the trained neural network model is used for monitoring.

9. The control method of the washing machine according to claim 7, characterized in that: The method for monitoring the eccentricity is: The eccentricity output by the second neural network model is used for monitoring; The second neural network model is trained by the following method: Acquire a data set, wherein the data set includes motor speed, total vector current Is, and total vector voltage Vs of the motor detected when several different eccentric loads are placed in the drum of the drum washing machine; The motor speed, the total vector current Is and the total vector voltage Vs of the motor are used as inputs of a second neural network model, wherein the output of the second neural network model includes eccentricity; The motor speed in the second neural network model data set is greater than the motor speed in the neural network model data set.

10. The control method of the washing machine according to claim 9, characterized in that: The formula of the jth neuron in the i-th layer of the hidden layer of the second neural network model is: ; in, is the output of the jth neuron node in the previous layer, for The weight of the jth neuron in the layer, for The bias of the jth neuron in the layer, i is the layer where the neuron is located, j is the number of neurons in each layer, for The total number of neurons in the layer; The output of the neuron is obtained through the activation function , activation function is ReLU; Mean absolute error loss function: ,in, is the actual value, is the output value of the neural network model, M is the number of samples, and m is the mth sample; Substitute the actual value and the output value of the neural network model into the loss function to calculate the loss value of the current neural network model; Using the chain rule starting from the output layer, calculate the gradient of the loss function with respect to each weight and bias. The layers are calculated as follows: ； in, For the layer weight vector, For the layer bias vector; According to the calculated gradient, the weights and biases of the second neural network model are updated: ； in, is the learning rate; The training is iterated until the loss function converges or the predetermined number of training rounds is reached, and a trained second neural network model is obtained.

Images