Washing machine control methods, devices and washing machines

Abstract

A washing machine control method, device, and washing machine are provided. The method includes: S1: Constructing a BP neural network model using the target motor speed, the weight of the clothes, and the motor frequency ramp-up rate as input layers, and the carrier ratio of the washing machine as the output layer; S2: Initializing the BP neural network and learning it through a learning mode; S3: After the learning mode ends, obtaining the carrier ratio calculated by the BP neural network model, and using the carrier ratio to control the washing machine motor when the obtained carrier ratio meets a preset threshold. This invention effectively reduces the power device losses while ensuring the safety of the switching devices.

Description

Technical Field

[0001] This invention relates to the field of automatic control, and more specifically to a washing machine control method, apparatus, and washing machine. Background Technology

[0002] As washing machine technology matures, inverter solutions are evolving towards lower losses and lower costs. To achieve higher washing ratios and more efficient washing methods, control technicians design washing cycles with higher spin speeds and frequency ramp-up rates. However, a constant carrier ratio during high-frequency motor operation leads to increased control system losses and current harmonics. Furthermore, prolonged use of excessively high frequencies causes severe overheating in power devices such as the IPM (Integrated Power Device), increasing component wear and reducing lifespan. Currently, there is no satisfactory solution to these problems.

[0003] Therefore, existing technologies require a solution that can effectively predict the carrier ratio and control the washing machine based on the carrier ratio.

[0004] The information disclosed in the background section above is only used to further understand the background of the present invention, and therefore may include information known to those skilled in the art that does not constitute prior art. Summary of the Invention

[0005] This invention relates to a washing machine control method, apparatus, and washing machine. According to the washing machine control scheme proposed by this invention, the carrier ratio of the motor is adjusted by the output of a BP neural network, effectively reducing the loss of power devices; furthermore, real-time control is achieved by using different ramp-up rates, target frequencies, and weighing results as inputs to the input layer of the BP neural network.

[0006] The first aspect of the present invention provides a washing machine control method, characterized in that the method includes: constructing a BP neural network model with the target motor speed of the washing machine, the weighing result of the clothes, and the motor frequency increase rate as the input layer and the carrier ratio of the washing machine as the output layer; initializing the BP neural network and learning it through a learning mode; after the learning mode ends, obtaining the carrier ratio calculated by the BP neural network model, and using the carrier ratio to control the motor of the washing machine when the obtained carrier ratio meets a preset threshold.

[0007] According to an embodiment of the present invention, the BP neural network model comprises an input layer, a hidden layer, and an output layer connected sequentially. In another embodiment, the input layer has 3 neurons, the hidden layer has 4 neurons, and the output layer has 1 neuron. During forward propagation of the BP neural network, the input of the BP neural network model is converted into an output through an activation function, and this output serves as the input to the next layer. During backward propagation of the BP neural network model, gradient descent is used to correct the weights and thresholds of each neuron, and least squares is used to represent the mean square error of the output. The mean square error is used to adjust the 12 weights between the input layer and the hidden layer, the 3 weights between the hidden layer and the output layer, the thresholds of the 4 neurons in the hidden layer, and the threshold of the 1 neuron in the output layer. The BP neural network model sets an iteration termination condition; when the training error meets the iteration termination condition, the BP neural network learning ends.

[0008] According to one embodiment of the present invention, the learning mode includes: calculating the outputs of the hidden layer and output layer units of the BP neural network model; adjusting the connection weights from the hidden layer to the output layer units and the threshold of the output layer; adjusting the connection weights from the input layer to the hidden layer and the threshold of the hidden layer; and updating the learning mode according to the adjustment results.

[0009] According to one embodiment of the present invention, the threshold of the output layer neuron and the threshold of the hidden layer neuron are related to the washing machine switching frequency and switching loss.

[0010] According to one embodiment of the present invention, in the learning mode of the BP neural network model, when the input layer motor up-frequency rate, target frequency, and weighing result are adjusted in an increasing manner, the threshold adjustment of the output layer carrier ratio is adjusted in a decreasing manner.

[0011] According to one embodiment of the present invention, the preset threshold is determined based on an acceptable upper limit for switching losses in the washing machine and a switching frequency.

[0012] According to an embodiment of the present invention, the activation function between the layers of the BP neural network model is a sigmoid logarithmic function tans ig.

[0013] According to an embodiment of the present invention, in step S3, when the carrier ratio calculated by the BP neural network model is obtained, the carrier ratio is normalized so that the preset threshold and the normalized carrier ratio can be compared.

[0014] According to an embodiment of the present invention, when the BP neural network model performs forward propagation, the operations performed are as follows: (1) calculating each neuron of the input layer and each neuron of the hidden layer by weighted summation; (2) calculating the neurons of the output layer using the hidden layer; (3) calculating the error using the output value and starting backward propagation.

[0015] According to an embodiment of the present invention, when the BP neural network model performs back propagation, the operation is as follows: (1) calculate the partial derivative of the error using the output result of forward propagation; (2) perform weighted summation using the partial derivative and the hidden layer until the input layer finally updates the weights using the partial derivative of each neuron.

[0016] According to one embodiment of the present invention, the partial derivative of the error is represented by the residual, and in the BP neural network model, the learning rate is the magnitude of each weight update.

[0017] According to one embodiment of the present invention, the formula for calculating the residual is as follows: from the output layer to the hidden layer: residual = -(output value - sample value) × derivative of the activation function; from the hidden layer to the hidden layer: residual = (weighted sum of the residuals of each neuron in the right layer) × derivative of the activation function; according to one or more embodiments of the present invention, the weight update formula of the BP neural network is as follows: from the input layer to the hidden layer: weight increase = input value × residual of the corresponding neuron in the right layer × learning rate; from the hidden layer to the output layer: weight increase = activation function of the current neuron × residual of the corresponding neuron in the right layer × learning rate; wherein the right layer is the right layer of the current layer when the BP neural network model propagates to the right.

[0018] According to one embodiment of the present invention, forward and backward propagation calculations are repeatedly performed on all data of the BP neural network model until the output error reaches a preset threshold, wherein the preset threshold can be a very small value.

[0019] According to one embodiment of the present invention, in the learning mode, the number of iterations is recorded when the BP neural network converges, and the number of iterations is the number of iterations in which the BP neural network does not get trapped in a local optimum.

[0020] According to one embodiment of the present invention, the BP neural network model aims to minimize the loss when training the weights, and when training the weights, it searches for a point where the error gradient is zero, and when calculating the gradient, it incorporates a random factor to determine when there are moments when the gradient is not zero even if the model gets stuck in a local minimum, so as to jump out of the local minimum and continue the search.

[0021] A second aspect of the present invention provides a washing machine control device, characterized in that it includes a memory and a processor; the memory is used to store a computer program; the processor is used to implement the washing machine control method according to the above when the computer program is executed.

[0022] A third aspect of the present invention provides a washing machine, characterized in that it uses the washing machine control method according to the above, or includes the washing machine control device according to the above.

[0023] This invention provides a method for adaptive carrier ratio adjustment by constructing the input of the output layer of a BP neural network using the motor's frequency ramp-up rate, target frequency, and weighing results, which can effectively reduce the losses of power devices. For certain special materials such as wool, designers may implement special designs in the washing cycle to protect the garments, such as designing low speed and high frequency ramp-up rate. This invention dynamically adjusts the carrier ratio based on these special frequency ramp-up rates, target frequencies, and weighing results to meet functional requirements while ensuring the safety of switching devices. Attached Figure Description

[0024] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the following description of the embodiments will be simplified. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 A schematic diagram of a BP neural network model according to an exemplary embodiment of the present invention is shown.

[0026] Figure 2 A flowchart of a washing machine control method according to an embodiment of the present invention is shown.

[0027] Figure 3 A schematic diagram of a constructed BP neural network model according to an exemplary implementation of the present invention is shown.

[0028] Figure 4 A schematic diagram of a constructed BP neural network model according to an exemplary implementation of the present invention is shown.

[0029] Figure 5 This is a training iteration convergence graph of a BP neural network model according to an exemplary embodiment of the present invention. Specific Implementation

[0030] As used herein, the terms "first," "second," etc., can be used to describe elements in exemplary embodiments of the present invention. These terms are used only to distinguish one element from another, and the inherent features or order of the corresponding elements are not limited by the term. Unless otherwise defined, all terms used herein (including technical or scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. Terms defined in common dictionaries are to be interpreted as having the same meaning as in the context of the relevant technical field, and not as having an ideal or overly formal meaning, unless explicitly defined as having such a meaning in this invention.

[0031] Those skilled in the art will understand that the apparatus and methods of the present invention described herein and illustrated in the accompanying drawings are non-limiting exemplary embodiments, and the scope of the invention is defined only by the claims. Features illustrated or described in conjunction with an exemplary embodiment may be combined with features of other embodiments. Such modifications and variations are included within the scope of the invention.

[0032] In the following description, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the drawings, detailed descriptions of known functions or configurations are omitted to avoid unnecessarily obscuring the key technical aspects of the invention. Furthermore, throughout the description, the same reference numerals always refer to the same circuits, modules, or units, and for the sake of brevity, repeated descriptions of the same circuits, modules, or units are omitted.

[0033] Furthermore, it should be understood that one or more of the following methods or aspects can be performed by at least one control unit or controller. The terms "control unit," "controller," "control module," or "main control module" can refer to a hardware device including a memory and a processor. The memory or computer-readable storage medium is configured to store program instructions, and the processor is specifically configured to execute the program instructions to perform one or more processes, which will be further described below. Moreover, it should be understood that, as those skilled in the art will recognize, the following methods can be performed by including a processor in conjunction with one or more other components.

[0034] Washing machines employ different rotation-to-stop ratios, accelerations, and target frequencies for different types of loads to ensure a high washing efficiency while minimizing clothing wear. However, in this specialized washing mode, if the carrier ratio remains constant, it can lead to either malfunctions in the motor control system or abnormal wear on the switching devices. Therefore, this invention employs a BP neural network model to address the problem of overheating in power devices such as the IPM due to carrier ratio mismatch during washing machine operation.

[0035] Backpropagation (BP) is one of the most important algorithms in deep learning neural networks. It is a multi-layer feedforward neural network trained using the backpropagation algorithm and is one of the most widely used neural network models. A BP neural network model includes: 1. Input layer: Input data; 2. Hidden layers: Data processing plants between the input and output layers, establishing connections between them through various parameters (weights, biases) and activation functions; 3. Output layer: Output result.

[0036] According to one or more embodiments of the present invention, a BP neural network involves forward propagation and backward propagation. Forward propagation functions as follows: neurons in the input layer receive various information from the outside world and transmit this information to neurons in the hidden layers. Intermediate hidden layer neurons process and transform the received information according to requirements. In practical applications, the intermediate hidden layers can be set as one or more hidden layer structures, and the information is transmitted to the output layer through the last hidden layer. This process is the forward propagation process of the BP neural network. Backward propagation functions as follows: when the error between the actual output and the ideal output exceeds the expectation, the error needs to be propagated backward. It starts from the output layer, and the error is corrected according to the gradient descent method, propagating sequentially to the hidden layers and the input layer. Through continuous forward propagation of information and backward propagation of error, the weights of each layer are continuously adjusted; this is the learning and training of the neural network. When the output error decreases to the expected level or the preset number of learning iterations, training ends, and the BP neural network completes its learning.

[0037] Figure 1 A schematic diagram of a BP neural network model according to an exemplary embodiment of the present invention is shown.

[0038] like Figure 1 The diagram illustrates the topology of a three-layer backpropagation (BP) network, comprising an input layer, an output layer, and a hidden layer (or sublayer). Each neuron is connected to all neurons in the next layer, but there are no connections between neurons within the same layer. Arrows represent the flow of information. Here, iWrRef represents the target motor speed, iWeight represents the weighing result of the clothes, iAccb represents the motor's frequency ramp-up rate, and the output layer represents the carrier ratio of the washing machine motor.

[0039] According to one or more embodiments of the present invention, the BP algorithm consists of two parts: forward propagation of information and backward propagation of error. During forward propagation, input information is calculated layer by layer from the input through the hidden layers to the output layer, and the state of each neuron only affects the state of the next layer of neurons. If the desired output is not obtained at the output layer, the error change value of the output layer is calculated, and then backward propagation is performed. The error signal is propagated back through the network along the original connection path to modify the weights of each neuron until the desired target is achieved.

[0040] According to one or more embodiments of the present invention, the number of neurons in the output layer of the BP neural network is 3 (target frequency, upsampling rate, weighing result), and the number of neurons in the hidden layer is 4. A sigmoid logarithmic function (tans ig) is selected as the activation function between layers. The input is converted into the output through the activation function, and this output signal is used as the input to the next layer. This completes one forward propagation of the BP neural network, achieving a nonlinear mapping between input and output. During backpropagation of the BP neural network, gradient descent is used to correct the weights and thresholds of each unit. The mean square error of the output is represented by the least squares method. The mean square error is used to adjust the 12 weights between the input layer and the hidden layer, the 3 weights between the hidden layer and the output layer, the thresholds of the 4 neurons in the hidden layer, and the threshold of the 1 neuron in the output layer. Finally, an iteration termination condition is set; the recursion terminates when the training error is less than a set value.

[0041] Figure 2 A flowchart of a washing machine control method according to an embodiment of the present invention is shown.

[0042] like Figure 2 As shown, in S1: a BP neural network model is constructed with the target speed of the washing machine motor, the weighing result of the clothes, and the motor frequency increase rate as the input layer and the carrier ratio of the washing machine as the output layer. The BP neural network model includes an input layer, a hidden layer, and an output layer connected in sequence.

[0043] In S2: the PB neural network is initialized and learned through the learning mode;

[0044] In S3: After the learning mode ends, the carrier ratio calculated by the BP neural network model is obtained. When the obtained carrier ratio meets the preset threshold, the carrier ratio is used to control the washing machine motor.

[0045] Figure 3 A schematic diagram of a constructed BP neural network model according to an exemplary implementation of the present invention is shown.

[0046] like Figure 3The BP neural network model includes an input layer, an output layer, and hidden layers, where W and b represent the weight parameters and bias parameters in the neural network.

[0047] According to one or more embodiments of the present invention, the neural network of the present invention is a processing system simulating the human brain, composed of a large number of processing units, and possesses many advantages, such as adaptability and nonlinearity. The BP neural network, or Back-Propagation Neural Network, is an artificial neural network adapted to nonlinear pattern recognition and classification pre-problems. BP networks have self-learning and adaptive characteristics, possess high nonlinearity and strong generalization ability, and can approximate nonlinear relationships with arbitrary precision, making them widely used in neural networks. It continuously adjusts the network weights and thresholds through training with sample data, causing the error function to decrease along the negative gradient direction, thus approximating the desired output.

[0048] like Figure 3 As shown, the BP neural network process consists of two stages: forward propagation and backward propagation. In the first stage, forward propagation, data is input from the input layer, multiplied by the corresponding weights along the network's path, summed, and then used as input in the activation function for calculation. The calculated result is then passed to the next neuron. This process continues sequentially from the input layer through the output layer until the final output is obtained, completing forward propagation. In the second stage, backward propagation, the output result is compared with the ideal output result, and the error between the output result and the ideal output result is propagated backward through the network. Specifically, this involves multiple iterations, continuously adjusting all weights between neurons (or nodes) on the network using gradient descent. In the BP neural network of this invention, each neuron has a threshold value; only when the stimulus received by a neuron exceeds the threshold will the neuron receive a stimulus, affecting the next neuron. During forward propagation of the BP neural network, the input of the BP neural network model is converted into the output through an activation function, and the output is used as the input of the next layer. During backward propagation of the BP neural network model, the weights and thresholds of each neuron are corrected using gradient descent, and the mean square error of the output is represented by the least squares method. The mean square error is used to adjust the 12 weights between the input layer and the hidden layer, the 3 weights between the hidden layer and the output layer, the thresholds of the 4 neurons in the hidden layer, and the threshold of the 1 neuron in the output layer.

[0049] According to one or more embodiments of the present invention, in a BP neural network model, a neuron only performs an action when its stimulation reaches a certain value, which is a threshold, i.e., the aforementioned limit. According to one or more embodiments of the present invention, the threshold can be adjusted by combining the switching frequency and switching losses, and determined based on hardware limitations and performance requirements through extensive experimentation.

[0050] According to one or more embodiments of the present invention, the operation performed during forward propagation of the BP neural network model is as follows:

[0051] (1) Calculate the input layer and the hidden layer by using a weighted summation method;

[0052] (2) Calculate the neurons in the output layer using the hidden layer;

[0053] (3) Calculate the error using the output value and start backpropagation.

[0054] According to one or more embodiments of the present invention, the operation performed during backpropagation of the BP neural network model is as follows:

[0055] (1) Calculate the partial derivative of the error using the output of forward propagation.

[0056] (2) Use partial derivatives and hidden layers to perform weighted summation until the input layer finally uses the partial derivatives of each neuron to update the weights;

[0057] According to one or more embodiments of the present invention, the partial derivative of the error is represented by the residual (error term). In the present invention, in the BP neural network model, the learning rate is the magnitude of each update.

[0058] According to one or more embodiments of the present invention, the formula for calculating the residual difference is as follows:

[0059] When moving from the output layer to the hidden layer: Residual = -(Output value - Sample value) × Derivative of the activation function;

[0060] When moving from one hidden layer to another: Residual = (weighted sum of the residuals of each neuron in the right layer) × derivative of the activation function;

[0061] According to one or more embodiments of the present invention, the weight update formula of the BP neural network is as follows:

[0062] From the input layer to the hidden layer: weight increase = input value × residual of the corresponding neuron in the right layer × learning rate;

[0063] From the hidden layer to the output layer: Weight increase = activation function of the current neuron × residual of the corresponding neuron in the right layer × learning rate;

[0064] Then, the above calculation is repeated for all the data of the BP neural network model until the output error reaches a preset threshold, where the preset threshold can be a very small value.

[0065] Figure 4 A schematic diagram of a constructed BP neural network model according to an exemplary implementation of the present invention is shown.

[0066] like Figure 4 As shown, (1) during the network initialization phase, the input data is determined. The input data of the input layer is the upsampling rate, the target frequency, and the weighing result; the output data is determined, and the carrier ratio N is determined.

[0067] (2) The learning mode of the input BP neural network can be obtained using Keras, an open-source artificial neural network library written in Python, according to one or more embodiments of the present invention. The purpose of learning is to adjust the weights and thresholds through continuous forward and backward propagation.

[0068] (3) Calculate the outputs of the hidden layers and the output layer. Select the sigmoid logarithm function tans ig as the activation function between layers. The input is transformed into the output through the activation function, and the output signal is used as the input of the next layer. Among them, the hidden layer is the hidden layer.

[0069] (4) Adjust the connection weights from the hidden layer to the output layer and the threshold of the output layer units. The connection weights are obtained through neural network training. The threshold refers to the limit at which a neuron will respond when stimulated. This threshold is also a limit obtained based on power electronics theory and experimental experience. For example, when the motor frequency increase rate is large, the target frequency is large, and the weighing result is heavy, the threshold of the output layer carrier ratio should be set smaller to ensure that the carrier ratio changes quickly enough in extreme cases. That is, in the learning mode of the BP neural network model, when the input layer motor frequency increase rate, target frequency, and weighing result are adjusted by increasing, the threshold of the output layer carrier ratio is adjusted by decreasing.

[0070] According to one or more embodiments of the present invention, for example, if a driver board can withstand an IPM (intelligent power module) efficiency of more than 90%, then the acceptable switching loss should be less than 10%. The switching frequency depends on the intersection of the modulation wave and the carrier wave. The higher the switching frequency, the better the smoothness of the current, but the greater the interference. Therefore, the threshold should be set in combination with the hardware feedback of the washing machine. If the upper limit of the acceptable switching loss is large and no unnecessary interference is generated, then a larger threshold should be selected.

[0071] (5) Adjust the parameters of the BP neural network based on the results of the above adjustments.

[0072] (6) Adjust the weights of each layer of the network according to certain rules until they converge to a certain value, and the learning mode ends. When the learning mode ends, obtain the carrier ratio output by the BP neural network model, and then perform a per-unit processing on the carrier ratio N for easy comparison with the upper limit value in the driving algorithm (i.e., Figure 4The middle box shows "Calculating the carrier ratio after learning". Typically, the standardization of physical quantities can be achieved by dividing the true value by the reference value, eliminating the unit dimension, and converting it into a value between 0 and 1 for easier comparison.

[0073] (7) When the carrier ratio is less than the preset upper limit, the training model of the BP neural network is terminated, and the carrier ratio is used to control the washing machine.

[0074] In particular, if the learning mode has not ended or the calculated carrier ratio is greater than the preset upper limit, the BP neural network model needs to re-enter the learning mode for learning and training.

[0075] According to one or more embodiments of the present invention, the selection of the ratio N of carrier frequency to modulation frequency has a great influence on the harmonics of the inverter output voltage. The larger N is, the lower the harmonic content. However, due to the limitations of actual switching devices, N cannot be too large.

[0076] Since the carrier ratio is the ratio of the fundamental wave (triangular wave) to the total number of pulses output by the sinusoidal modulating wave in modulation, i.e. the ratio of their frequencies fv / fs, the carrier ratio output value is determined by both the acceptable upper limit of the switching transistor loss and the switching frequency. Therefore, the prediction model selects the sigmoid logarithmic function tans ig as the activation function of the output layer neurons.

[0077] In reality, the load changes of a washing machine are not a linear system, and its real-time, accurate mathematical model is difficult to obtain. Therefore, it is challenging to accurately express and effectively control the dynamic changes of the washing machine's drive motor using classical control theory. However, the BP neural network algorithm does not require a mathematical model of the controlled object, can effectively control the dynamic performance of the system, and exhibits good robustness. For any washing process with different ramp-up rates and target speeds, and with a random load, the BP neural network algorithm uses the target speed, weighing result, and ramp-up rate as inputs to the output layer of the BP neural network to achieve adaptive carrier ratio adjustment. As described above, using the BP neural network algorithm to achieve adaptive carrier ratio adjustment can not only effectively reduce the switching frequency of power devices, reduce losses, and suppress current harmonics, but also effectively reduce noise and enhance the anti-eccentricity capability of the pulsator washing machine.

[0078] Figure 5 This is a training iteration convergence graph of a BP neural network model according to an exemplary embodiment of the present invention.

[0079] like Figure 5As shown in the training graph, the neural network converges after approximately 137 iterations. Although the BP neural network performs well, the excessive number of iterations makes it prone to getting trapped in local optima. Therefore, in the learning mode described above, the number of iterations is recorded when the BP neural network converges. This number of iterations represents the number of iterations required for the BP neural network to avoid getting trapped in local optima, or the number of iterations required to achieve optimal performance. The neural network model of this invention may have many local optima rather than a global optimum; local minima represent local optima.

[0080] According to one or more embodiments of the present invention, the BP neural network model aims to minimize the loss when training the weights, and when training the weights, it searches for a point where the error gradient is 0, and when calculating the gradient, it incorporates a random factor to determine when there are moments when the gradient is not zero even if the model gets stuck in a local minimum, so as to jump out of the local minimum and continue the search.

[0081] The present invention also provides a washing machine control device, including a memory and a processor; the memory is used to store a computer program; the processor is used to implement the above-described washing machine control method when the computer program is executed.

[0082] The present invention also provides a washing machine that uses the washing machine control method described above, or includes the washing machine control device described above.

[0083] The present invention also provides a compressor comprising the permanent magnet synchronous motor according to the above.

[0084] According to one or more embodiments of the present invention, the control logic of the present invention can implement the processing of the processes in the system described above using encoded instructions (e.g., computer and / or machine-readable instructions) stored on a non-transitory computer and / or machine-readable medium (e.g., hard disk drive, flash memory, read-only memory, optical disk, digital multifunction disk, cache, random access memory, and / or any other storage device or storage disk), storing information for any time period (e.g., extended time periods, permanent, transient instances, temporary caches, and / or information caches) in the non-transitory computer and / or machine-readable medium. As used herein, the term "non-transitory computer-readable medium" is explicitly defined to include any type of computer-readable storage device and / or storage disk, excluding propagated signals and transmission media.

[0085] According to one or more embodiments of the present invention, the logic in the system of the present invention can be implemented using control circuitry (control logic, main control system, or control module), which may include one or more processors and may also internally include non-transitory computer-readable media. Specifically, the main control system or control module may include a microcontroller (MCU). The processor used to implement the processing of the logic in the system of the present invention may be, for example, but not limited to, one or more single-core or multi-core processors. The processor(s) may include any combination of general-purpose processors and special-purpose processors (e.g., graphics processors, application processors, etc.). The processor may be coupled thereto and / or may include memory / storage devices and may be configured to execute instructions stored in the memory / storage devices to implement various applications and / or operating systems running on the controller in the present invention.

[0086] The accompanying drawings and detailed description of the invention, cited above as examples, serve to explain the invention but do not limit its meaning or scope as described in the claims. Therefore, those skilled in the art can readily make modifications from the above description. Furthermore, those skilled in the art can remove some of the components described herein without degrading performance, or add other components to improve performance. Additionally, those skilled in the art can change the order of steps in the method described herein depending on the process or equipment environment. Therefore, the scope of the invention should not be determined by the embodiments described above, but rather by the claims and their equivalents.

[0087] Although the invention has been described in conjunction with embodiments now considered to be achievable, it should be understood that the invention is not limited to the disclosed embodiments, but rather is intended to cover various modifications and equivalent configurations included within the spirit and scope of the appended claims.

Claims

1. A washing machine control method, characterized in that, The method includes: A BP neural network model is constructed using the target speed of the washing machine motor, the weight of the clothes, and the motor frequency increase rate as the input layer, and the carrier ratio of the washing machine as the output layer. The BP neural network model includes an input layer, a hidden layer, and an output layer connected in sequence. The BP neural network is initialized and then learned through a learning mode; After the learning mode ends, the carrier ratio calculated by the BP neural network model is obtained. When the obtained carrier ratio meets the preset threshold, the carrier ratio is used to control the washing machine motor. The learning modes include: During forward propagation of the BP neural network model, the following operations are performed: each neuron in the input layer and each neuron in the hidden layer are calculated using a weighted summation method; the neurons in the output layer are calculated using the hidden layers; the error is calculated using the output value, and backpropagation begins. When the BP neural network model performs backpropagation, the following operations are performed: the partial derivatives of the error are calculated using the output of the forward propagation; the partial derivatives and the hidden layers are weighted and summed until the input layer finally updates the weights using the partial derivatives of each neuron; The forward and backward propagation calculations are repeatedly performed on all the data of the BP neural network model until the output error reaches a preset threshold condition.

2. The method according to claim 1, characterized in that, The residual is used to represent the partial derivative of the error, and in the BP neural network model, the learning rate is used to represent the magnitude of each weight update. The formula for calculating the residual in the BP neural network model is as follows: When moving from the output layer to the hidden layer: Residual = -(Output value - Sample value) × Derivative of the activation function; When moving from one hidden layer to another: Residual = (weighted sum of the residuals of each neuron in the right layer) × derivative of the activation function; The weight update formula for the BP neural network model is as follows: From the input layer to the hidden layer: weight increase = input value × residual of the corresponding neuron in the right layer × learning rate; From the hidden layer to the output layer: Weight increase = activation function of the current neuron × residual of the corresponding neuron in the right layer × learning rate; The right layer here refers to the right layer of the current layer when the BP neural network model propagates to the right.

3. The method according to claim 1, characterized in that, The threshold values ​​of the output layer neurons and the hidden layer neurons are related to the washing machine's switching frequency and switching losses.

4. The method according to claim 1, characterized in that, in, In the learning mode of the BP neural network model, when the input layer motor frequency increase rate, target frequency, and weighing result are adjusted in an increasing manner, the output layer carrier ratio threshold is adjusted in a decreasing manner.

5. The method according to claim 1, characterized in that, in, The preset threshold is determined based on the acceptable upper limit of switching losses and the switching frequency in the washing machine.

6. The method according to claim 1, characterized in that, The activation function between the layers of the BP neural network model is the sigmoid logarithmic function tansig.

7. The method according to claim 1, characterized in that, in, In the learning mode, the number of iterations is recorded when the BP neural network converges. The number of iterations is the number of iterations in which the BP neural network does not get trapped in a local optimum.

8. The method according to claim 7, characterized in that, The BP neural network model aims to minimize the loss when training the weights. During weight training, it searches for the point where the error gradient is zero. When calculating the gradient, a random factor is added to determine the moment when the gradient is not zero even if the model gets stuck in a local minimum, so that it can jump out of the local minimum and continue the search.

9. A washing machine control device, characterized in that, It includes a memory and a processor; the memory is used to store a computer program; the processor is used to implement the washing machine control method according to any one of claims 1 to 8 when the computer program is executed.

10. A washing machine, characterized in that, It uses the washing machine control method according to any one of claims 1-8, or includes the washing machine control device according to any one of claims 9.

Images